{"id":2789822,"date":"2025-05-12T11:31:52","date_gmt":"2025-05-12T18:31:52","guid":{"rendered":"https:\/\/www.esri.com\/arcgis-blog\/?post_type=blog&p=2789822"},"modified":"2025-05-16T18:15:47","modified_gmt":"2025-05-17T01:15:47","slug":"task-and-generalized-pretrained-models","status":"publish","type":"blog","link":"https:\/\/www.esri.com\/arcgis-blog\/products\/arcgis\/geoai\/task-and-generalized-pretrained-models","title":{"rendered":"Pretrained Models in ArcGIS: Comparing Task-Specific and Generalized Vision-Language models"},"author":8452,"featured_media":0,"parent":0,"menu_order":0,"comment_status":"open","ping_status":"closed","template":"","format":"standard","meta":{"_acf_changed":false,"_searchwp_excluded":""},"categories":[770712,22931],"tags":[779412,186132,665211,758872],"industry":[],"product":[421922,36841,36951,36581,36551,36561],"class_list":["post-2789822","blog","type-blog","status-publish","format-standard","hentry","category-geoai","category-imagery","tag-arcgis-geoai","tag-deep-learning","tag-geoai","tag-pretrained-models","product-arcgis","product-api-python","product-image-server","product-arcgis-living-atlas","product-arcgis-online","product-arcgis-pro"],"acf":{"authors":[{"ID":8452,"user_firstname":"Vinay","user_lastname":"Viswambharan","nickname":"Vinay Viswambharan","user_nicename":"vinayv","display_name":"Vinay Viswambharan","user_email":"vinayv@esri.com","user_url":"https:\/\/www.esri.com\/arcgis-blog\/author\/vinayv\/","user_registered":"2018-10-04 22:28:54","user_description":"Principal Product manager on the Imagery team at Esri, with a zeal for remote sensing, AI and everything imagery.","user_avatar":""},{"ID":6911,"user_firstname":"Rohit","user_lastname":"Singh","nickname":"Rohit Singh","user_nicename":"rsinghesri-com","display_name":"Rohit Singh","user_email":"rsingh@esri.com","user_url":"","user_registered":"2018-03-02 00:19:00","user_description":"Rohit Singh is Director of Esri\u2019s R&D Center in New Delhi, leading the design and development of Geospatial AI capabilities across the ArcGIS platform. He has played a key role in the development of ArcGIS API for Python, ArcGIS Java Engine API, and the Linux enablement of ArcGIS. An alumnus of IIT Kharagpur, Rohit holds an MS in Computer Science with specialization in AI from Georgia Tech.","user_avatar":""}],"short_description":"Explore ArcGIS AI models\u2014compare precise task models vs. flexible vision-language tools, and see how Esri makes AI easy for all.","flexible_content":[{"acf_fc_layout":"content","content":"

ArcGIS provides nearly 100 pre-trained AI models<\/a>, making geospatial AI more accessible than ever before. These models are part of Esri\u2019s ongoing effort to democratize AI, helping users\u2014regardless of their technical background\u2014extract high-quality insights from imagery with minimal investment in compute or skillset.<\/p>\n

For some users, these models are a starting point into the world of AI. For others, they are a complete solution, solving domain-specific problems right out of the box or serving as a foundation for fine-tuning to specific geographies, datasets or features of interest.<\/p>\n

ArcGIS is democratizing access to AI by providing pre-trained models that deliver authoritative geospatial insights\u2014making AI approachable, scalable, and impactful for a wide range of users in the geospatial community.<\/p>\n

Two Types of Pretrained Models<\/h1>\n

Task-Specific Pretrained Models<\/strong> are built to do one job extremely well\u2014like detecting building footprints, identifying swimming pools, or classifying land cover. They are highly optimized for single tasks and perform well in specific domains and imagery types.<\/p>\n

Examples include models for building footprint extraction,<\/a> land cover classification<\/a>, tree segmentation<\/a>, and car detection<\/a>.<\/p>\n

Generalized Vision-Language Pretrained Models<\/strong>, on the other hand, are like Swiss Army knives. They can perform many visual tasks using natural language prompts\u2014such as \u201cfind red cars,\u201d \u201csegment all lakes,\u201d or \u201cclassify these images based on candidate labels\u201d\u2014all with the same model.<\/p>\n

Examples include Text SAM<\/a>, Prompt-Based Segmentation<\/a>, Vision-Language Context-Based Classification<\/a>, and Image Interrogation<\/a>.<\/p>\n"},{"acf_fc_layout":"content","content":"

Key Differences Between the Two Model Types<\/h1>\n

The models can be differentiated based on 7 factors:<\/p>\n

    \n
  1. Applications<\/a> – Refers to the types of real-world use cases the model is designed to support- such as extracting features, classifying objects, or answering questions about imagery. This defines whether the model serves a narrow, fixed purpose or can adapt to diverse user needs.<\/li>\n
  2. Accuracy<\/a> – Measures how precisely and reliably a model performs its intended task. This includes how well it detects, classifies, or segments features, and how often it makes correct predictions based on ground truth.<\/li>\n
  3. Supported Data Types<\/a> – Describes the kinds of imagery or input formats the model can work with\u2014such as satellite imagery, drone data, street-level photos, or point cloud datasets. It also reflects the model\u2019s compatibility with different resolutions and spectral bands.<\/li>\n
  4. Tasks<\/a> – Defines the range and nature of operations the model can perform\u2014like object detection, pixel classification, image captioning, or point cloud classification. It helps distinguish between models that do one task well and those that can handle many.<\/li>\n
  5. Model Size<\/a> – Refers to the computational footprint of the model, including the number of parameters, storage requirements, and memory usage. Larger models often support more complex tasks but require more resources to run.<\/li>\n
  6. Fine-tuning<\/a> – Indicates how easily the model can be adapted to specific datasets, geographies, or new classes. Some models need retraining for different use cases, while others can generalize with minimal or no customization.<\/li>\n
  7. Domain<\/a> – Represents the scope of subject matter the model is trained on\u2014such as remote sensing imagery or general photos or Alt-text pairs from the Internetimagery. A model\u2019s domain influences its performance and relevance for geospatial applications.<\/li>\n<\/ol>\n

     <\/p>\n

    1. Applications<\/h2>\n

    Task-specific models are built for a single, well-defined task like detecting buildings or swimming pools. In contrast, generalized models can handle many tasks through flexible natural language prompts, such as detecting red cars or airplanes.<\/p>\n"},{"acf_fc_layout":"image","image":{"ID":2790252,"id":2790252,"title":"1-Applications","filename":"1-Applications.jpg","filesize":231652,"url":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/1-Applications.jpg","link":"https:\/\/www.esri.com\/arcgis-blog\/products\/arcgis\/geoai\/task-and-generalized-pretrained-models\/1-applications","alt":"","author":"8452","description":"Applications","caption":"Applications","name":"1-applications","status":"inherit","uploaded_to":2789822,"date":"2025-05-12 18:30:53","modified":"2025-05-12 18:31:12","menu_order":0,"mime_type":"image\/jpeg","type":"image","subtype":"jpeg","icon":"https:\/\/www.esri.com\/arcgis-blog\/wp-includes\/images\/media\/default.png","width":1534,"height":922,"sizes":{"thumbnail":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/1-Applications-213x200.jpg","thumbnail-width":213,"thumbnail-height":200,"medium":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/1-Applications.jpg","medium-width":434,"medium-height":261,"medium_large":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/1-Applications.jpg","medium_large-width":768,"medium_large-height":462,"large":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/1-Applications.jpg","large-width":1534,"large-height":922,"1536x1536":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/1-Applications.jpg","1536x1536-width":1534,"1536x1536-height":922,"2048x2048":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/1-Applications.jpg","2048x2048-width":1534,"2048x2048-height":922,"card_image":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/1-Applications-774x465.jpg","card_image-width":774,"card_image-height":465,"wide_image":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/1-Applications.jpg","wide_image-width":1534,"wide_image-height":922}},"image_position":"center","orientation":"horizontal","hyperlink":""},{"acf_fc_layout":"content","content":"

    2. Accuracy<\/h2>\n

    Task-specific models usually provide higher accuracy since they’re fine-tuned for a specific job. Generalized models offer broader capabilities but may compromise on precision.<\/p>\n"},{"acf_fc_layout":"image","image":{"ID":2792792,"id":2792792,"title":"2-accuracy","filename":"2-accuracy.jpg","filesize":79516,"url":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/2-accuracy.jpg","link":"https:\/\/www.esri.com\/arcgis-blog\/products\/arcgis\/geoai\/task-and-generalized-pretrained-models\/2-accuracy","alt":"accuracy of model","author":"6911","description":"accuracy of model","caption":"Accuracy of model","name":"2-accuracy","status":"inherit","uploaded_to":2789822,"date":"2025-05-14 10:01:01","modified":"2025-05-14 10:01:20","menu_order":0,"mime_type":"image\/jpeg","type":"image","subtype":"jpeg","icon":"https:\/\/www.esri.com\/arcgis-blog\/wp-includes\/images\/media\/default.png","width":1486,"height":605,"sizes":{"thumbnail":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/2-accuracy-213x200.jpg","thumbnail-width":213,"thumbnail-height":200,"medium":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/2-accuracy.jpg","medium-width":464,"medium-height":189,"medium_large":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/2-accuracy.jpg","medium_large-width":768,"medium_large-height":313,"large":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/2-accuracy.jpg","large-width":1486,"large-height":605,"1536x1536":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/2-accuracy.jpg","1536x1536-width":1486,"1536x1536-height":605,"2048x2048":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/2-accuracy.jpg","2048x2048-width":1486,"2048x2048-height":605,"card_image":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/2-accuracy-826x336.jpg","card_image-width":826,"card_image-height":336,"wide_image":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/2-accuracy.jpg","wide_image-width":1486,"wide_image-height":605}},"image_position":"center","orientation":"horizontal","hyperlink":""},{"acf_fc_layout":"content","content":"

    3. Supported Data Types<\/h2>\n

    Task-specific models in ArcGIS are versatile in their input support and support a wide variety of geospatial formats, including low-resolution satellite imagery, NAIP, 3D point clouds, SAR, hyperspectral, and video. Generalized models typically work only with RGB imagery.<\/p>\n"},{"acf_fc_layout":"image","image":{"ID":2790282,"id":2790282,"title":"3-Supported-Data-Types","filename":"3-Supported-Data-Types.jpg","filesize":310287,"url":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/3-Supported-Data-Types.jpg","link":"https:\/\/www.esri.com\/arcgis-blog\/products\/arcgis\/geoai\/task-and-generalized-pretrained-models\/3-supported-data-types","alt":"Supported data types","author":"8452","description":"Supported data types","caption":"Supported data types","name":"3-supported-data-types","status":"inherit","uploaded_to":2789822,"date":"2025-05-12 18:43:23","modified":"2025-05-12 18:43:38","menu_order":0,"mime_type":"image\/jpeg","type":"image","subtype":"jpeg","icon":"https:\/\/www.esri.com\/arcgis-blog\/wp-includes\/images\/media\/default.png","width":1541,"height":927,"sizes":{"thumbnail":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/3-Supported-Data-Types-213x200.jpg","thumbnail-width":213,"thumbnail-height":200,"medium":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/3-Supported-Data-Types.jpg","medium-width":434,"medium-height":261,"medium_large":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/3-Supported-Data-Types.jpg","medium_large-width":768,"medium_large-height":462,"large":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/3-Supported-Data-Types.jpg","large-width":1541,"large-height":927,"1536x1536":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/3-Supported-Data-Types-1536x924.jpg","1536x1536-width":1536,"1536x1536-height":924,"2048x2048":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/3-Supported-Data-Types.jpg","2048x2048-width":1541,"2048x2048-height":927,"card_image":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/3-Supported-Data-Types-773x465.jpg","card_image-width":773,"card_image-height":465,"wide_image":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/3-Supported-Data-Types.jpg","wide_image-width":1541,"wide_image-height":927}},"image_position":"center","orientation":"horizontal","hyperlink":""},{"acf_fc_layout":"content","content":"

    4. Tasks<\/h2>\n

    Task-specific models can handle a range of geospatial AI tasks: object detection, pixel classification, change detection, 3D point cloud classification, and more. Generalized models are limited to classification, segmentation, and object detection on RGB inputs.<\/p>\n"},{"acf_fc_layout":"image","image":{"ID":2790402,"id":2790402,"title":"4a-Tasks","filename":"4a-Tasks.jpg","filesize":267550,"url":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/4a-Tasks.jpg","link":"https:\/\/www.esri.com\/arcgis-blog\/products\/arcgis\/geoai\/task-and-generalized-pretrained-models\/4a-tasks","alt":"Supported tasks","author":"8452","description":"Supported tasks","caption":"Supported tasks","name":"4a-tasks","status":"inherit","uploaded_to":2789822,"date":"2025-05-12 19:44:03","modified":"2025-05-12 19:44:35","menu_order":0,"mime_type":"image\/jpeg","type":"image","subtype":"jpeg","icon":"https:\/\/www.esri.com\/arcgis-blog\/wp-includes\/images\/media\/default.png","width":1571,"height":929,"sizes":{"thumbnail":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/4a-Tasks-213x200.jpg","thumbnail-width":213,"thumbnail-height":200,"medium":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/4a-Tasks.jpg","medium-width":441,"medium-height":261,"medium_large":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/4a-Tasks.jpg","medium_large-width":768,"medium_large-height":454,"large":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/4a-Tasks.jpg","large-width":1571,"large-height":929,"1536x1536":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/4a-Tasks-1536x908.jpg","1536x1536-width":1536,"1536x1536-height":908,"2048x2048":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/4a-Tasks.jpg","2048x2048-width":1571,"2048x2048-height":929,"card_image":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/4a-Tasks-786x465.jpg","card_image-width":786,"card_image-height":465,"wide_image":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/4a-Tasks.jpg","wide_image-width":1571,"wide_image-height":929}},"image_position":"center","orientation":"horizontal","hyperlink":""},{"acf_fc_layout":"content","content":"

    5. Model Size<\/h2>\n

    Task-specific models are smaller and more efficient, making them easier to deploy and run with lower compute requirements. Generalized models tend to be large and compute-intensive, requiring more powerful GPUs or cloud services.<\/p>\n"},{"acf_fc_layout":"image","image":{"ID":2790372,"id":2790372,"title":"5-ModelSize","filename":"5-ModelSize.jpg","filesize":172744,"url":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/5-ModelSize.jpg","link":"https:\/\/www.esri.com\/arcgis-blog\/products\/arcgis\/geoai\/task-and-generalized-pretrained-models\/5-modelsize","alt":"Model size","author":"8452","description":"Model size","caption":"Model size","name":"5-modelsize","status":"inherit","uploaded_to":2789822,"date":"2025-05-12 19:39:28","modified":"2025-05-12 19:39:49","menu_order":0,"mime_type":"image\/jpeg","type":"image","subtype":"jpeg","icon":"https:\/\/www.esri.com\/arcgis-blog\/wp-includes\/images\/media\/default.png","width":1540,"height":842,"sizes":{"thumbnail":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/5-ModelSize-213x200.jpg","thumbnail-width":213,"thumbnail-height":200,"medium":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/5-ModelSize.jpg","medium-width":464,"medium-height":254,"medium_large":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/5-ModelSize.jpg","medium_large-width":768,"medium_large-height":420,"large":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/5-ModelSize.jpg","large-width":1540,"large-height":842,"1536x1536":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/5-ModelSize-1536x840.jpg","1536x1536-width":1536,"1536x1536-height":840,"2048x2048":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/5-ModelSize.jpg","2048x2048-width":1540,"2048x2048-height":842,"card_image":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/5-ModelSize-826x452.jpg","card_image-width":826,"card_image-height":452,"wide_image":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/5-ModelSize.jpg","wide_image-width":1540,"wide_image-height":842}},"image_position":"center","orientation":"horizontal","hyperlink":""},{"acf_fc_layout":"content","content":"

    6. Fine-tuning<\/h2>\n

    Task-specific models can be fine-tuned using a relatively small amount of labeled data to improve accuracy. Generalized models are typically static and not fine-tunable or improved.<\/p>\n"},{"acf_fc_layout":"image","image":{"ID":2790422,"id":2790422,"title":"6-fine-tunable","filename":"6-fine-tunable.jpg","filesize":111501,"url":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/6-fine-tunable.jpg","link":"https:\/\/www.esri.com\/arcgis-blog\/products\/arcgis\/geoai\/task-and-generalized-pretrained-models\/6-fine-tunable","alt":"Fine-tuning","author":"8452","description":"Fine-tuning","caption":"Fine-tuning","name":"6-fine-tunable","status":"inherit","uploaded_to":2789822,"date":"2025-05-12 20:02:30","modified":"2025-05-12 20:02:44","menu_order":0,"mime_type":"image\/jpeg","type":"image","subtype":"jpeg","icon":"https:\/\/www.esri.com\/arcgis-blog\/wp-includes\/images\/media\/default.png","width":1463,"height":1020,"sizes":{"thumbnail":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/6-fine-tunable-213x200.jpg","thumbnail-width":213,"thumbnail-height":200,"medium":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/6-fine-tunable.jpg","medium-width":374,"medium-height":261,"medium_large":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/6-fine-tunable.jpg","medium_large-width":768,"medium_large-height":535,"large":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/6-fine-tunable.jpg","large-width":1463,"large-height":1020,"1536x1536":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/6-fine-tunable.jpg","1536x1536-width":1463,"1536x1536-height":1020,"2048x2048":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/6-fine-tunable.jpg","2048x2048-width":1463,"2048x2048-height":1020,"card_image":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/6-fine-tunable-667x465.jpg","card_image-width":667,"card_image-height":465,"wide_image":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/6-fine-tunable.jpg","wide_image-width":1463,"wide_image-height":1020}},"image_position":"center","orientation":"horizontal","hyperlink":""},{"acf_fc_layout":"content","content":"

    7. Domain<\/h2>\n

    Task-specific models are purpose-built for geospatial analysis and closely aligned with real-world mapping needs. Generalized models are designed for broad, non-spatial image understanding tasks. By integrating them with ArcGIS, we can extend their capabilities to spatial domains, enabling powerful workflows that combine foundational visual understanding with location intelligence<\/p>\n"},{"acf_fc_layout":"image","image":{"ID":2790412,"id":2790412,"title":"7-Domain","filename":"7-Domain.jpg","filesize":263279,"url":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/7-Domain.jpg","link":"https:\/\/www.esri.com\/arcgis-blog\/products\/arcgis\/geoai\/task-and-generalized-pretrained-models\/7-domain","alt":"Domain","author":"8452","description":"Domain","caption":"Domain","name":"7-domain","status":"inherit","uploaded_to":2789822,"date":"2025-05-12 20:01:52","modified":"2025-05-12 20:02:04","menu_order":0,"mime_type":"image\/jpeg","type":"image","subtype":"jpeg","icon":"https:\/\/www.esri.com\/arcgis-blog\/wp-includes\/images\/media\/default.png","width":1516,"height":802,"sizes":{"thumbnail":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/7-Domain-213x200.jpg","thumbnail-width":213,"thumbnail-height":200,"medium":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/7-Domain.jpg","medium-width":464,"medium-height":245,"medium_large":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/7-Domain.jpg","medium_large-width":768,"medium_large-height":406,"large":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/7-Domain.jpg","large-width":1516,"large-height":802,"1536x1536":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/7-Domain.jpg","1536x1536-width":1516,"1536x1536-height":802,"2048x2048":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/7-Domain.jpg","2048x2048-width":1516,"2048x2048-height":802,"card_image":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/7-Domain-826x437.jpg","card_image-width":826,"card_image-height":437,"wide_image":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2025\/05\/7-Domain.jpg","wide_image-width":1516,"wide_image-height":802}},"image_position":"center","orientation":"horizontal","hyperlink":""},{"acf_fc_layout":"content","content":"

    Wrapping Up<\/h1>\n

    Both types of models have their strengths. Task-specific models are ideal for users who need high precision and domain-specific capabilities. Generalized models are great for exploring flexible, prompt-driven AI tasks. Our recommendation is to use a task-specific model if one exists for your features of interest and geography. If that doesn\u2019t exist, you can try using vision-language models and try out different prompts to see which ones give the best results. If you\u2019re still not getting the desired results, you can look into fine-tuning the task-specific models on your specific imagery, features of interest, and geography.<\/p>\n

    At Esri, we\u2019re committed to supporting both\u2014so our users can get the most out of imagery AI, no matter their experience level.<\/p>\n"}],"related_articles":[{"ID":2192712,"post_author":"350252","post_date":"2023-12-22 11:32:43","post_date_gmt":"2023-12-22 19:32:43","post_content":"","post_title":"ArcGIS AI Models \u2013 Year in Review","post_excerpt":"","post_status":"publish","comment_status":"open","ping_status":"closed","post_password":"","post_name":"esri-arcgis-ai-models-year-in-review","to_ping":"","pinged":"","post_modified":"2024-01-27 20:32:05","post_modified_gmt":"2024-01-28 04:32:05","post_content_filtered":"","post_parent":0,"guid":"https:\/\/www.esri.com\/arcgis-blog\/?post_type=blog&p=2192712","menu_order":0,"post_type":"blog","post_mime_type":"","comment_count":"0","filter":"raw"},{"ID":2515452,"post_author":"348302","post_date":"2024-10-10 07:00:37","post_date_gmt":"2024-10-10 14:00:37","post_content":"","post_title":"New Pretrained Geospatial AI Models for Disaster Response","post_excerpt":"","post_status":"publish","comment_status":"open","ping_status":"closed","post_password":"","post_name":"new-pretrained-geospatial-ai-models-for-disaster-response","to_ping":"","pinged":"","post_modified":"2024-10-10 09:24:23","post_modified_gmt":"2024-10-10 16:24:23","post_content_filtered":"","post_parent":0,"guid":"https:\/\/www.esri.com\/arcgis-blog\/?post_type=blog&p=2515452","menu_order":0,"post_type":"blog","post_mime_type":"","comment_count":"0","filter":"raw"},{"ID":2645732,"post_author":"366412","post_date":"2025-02-11 15:50:28","post_date_gmt":"2025-02-11 23:50:28","post_content":"","post_title":"Learn to use GeoAI to extract information from World Imagery","post_excerpt":"","post_status":"publish","comment_status":"closed","ping_status":"closed","post_password":"","post_name":"learn-to-use-ai-to-extract-information-from-world-imagery","to_ping":"","pinged":"","post_modified":"2025-08-05 11:31:58","post_modified_gmt":"2025-08-05 18:31:58","post_content_filtered":"","post_parent":0,"guid":"https:\/\/www.esri.com\/arcgis-blog\/?post_type=blog&p=2645732","menu_order":0,"post_type":"blog","post_mime_type":"","comment_count":"0","filter":"raw"}],"show_article_image":false,"card_image":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2023\/12\/banners_icon.jpg","wide_image":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2023\/12\/banner.gif"},"yoast_head":"\nTask vs Generalized Vision Models in ArcGIS<\/title>\n<meta name=\"description\" content=\"Explore the differences between task-specific and generalized pretrained models in ArcGIS for enhanced vision-language AI applications\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/www.esri.com\/arcgis-blog\/products\/arcgis\/geoai\/task-and-generalized-pretrained-models\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Pretrained Models in ArcGIS: Comparing Task-Specific and Generalized Vision-Language models\" \/>\n<meta property=\"og:description\" content=\"Explore the differences between task-specific and generalized pretrained models in ArcGIS for enhanced vision-language AI applications\" \/>\n<meta property=\"og:url\" content=\"https:\/\/www.esri.com\/arcgis-blog\/products\/arcgis\/geoai\/task-and-generalized-pretrained-models\" \/>\n<meta property=\"og:site_name\" content=\"ArcGIS Blog\" \/>\n<meta property=\"article:publisher\" content=\"https:\/\/www.facebook.com\/esrigis\/\" \/>\n<meta property=\"article:modified_time\" content=\"2025-05-17T01:15:47+00:00\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:site\" content=\"@ESRI\" \/>\n<meta name=\"twitter:label1\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data1\" content=\"6 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":[\"Article\",\"BlogPosting\"],\"@id\":\"https:\/\/www.esri.com\/arcgis-blog\/products\/arcgis\/geoai\/task-and-generalized-pretrained-models#article\",\"isPartOf\":{\"@id\":\"https:\/\/www.esri.com\/arcgis-blog\/products\/arcgis\/geoai\/task-and-generalized-pretrained-models\"},\"author\":{\"name\":\"Vinay Viswambharan\",\"@id\":\"https:\/\/www.esri.com\/arcgis-blog\/#\/schema\/person\/fe6ac8773de48adda0651f5f144d2acf\"},\"headline\":\"Pretrained Models in ArcGIS: Comparing Task-Specific and Generalized Vision-Language models\",\"datePublished\":\"2025-05-12T18:31:52+00:00\",\"dateModified\":\"2025-05-17T01:15:47+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\/\/www.esri.com\/arcgis-blog\/products\/arcgis\/geoai\/task-and-generalized-pretrained-models\"},\"wordCount\":10,\"commentCount\":0,\"publisher\":{\"@id\":\"https:\/\/www.esri.com\/arcgis-blog\/#organization\"},\"keywords\":[\"arcgis geoai\",\"deep learning\",\"geoAI\",\"pretrained models\"],\"articleSection\":[\"AI\",\"Imagery & Remote Sensing\"],\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"CommentAction\",\"name\":\"Comment\",\"target\":[\"https:\/\/www.esri.com\/arcgis-blog\/products\/arcgis\/geoai\/task-and-generalized-pretrained-models#respond\"]}]},{\"@type\":\"WebPage\",\"@id\":\"https:\/\/www.esri.com\/arcgis-blog\/products\/arcgis\/geoai\/task-and-generalized-pretrained-models\",\"url\":\"https:\/\/www.esri.com\/arcgis-blog\/products\/arcgis\/geoai\/task-and-generalized-pretrained-models\",\"name\":\"Task vs Generalized Vision Models in ArcGIS\",\"isPartOf\":{\"@id\":\"https:\/\/www.esri.com\/arcgis-blog\/#website\"},\"datePublished\":\"2025-05-12T18:31:52+00:00\",\"dateModified\":\"2025-05-17T01:15:47+00:00\",\"description\":\"Explore the differences between task-specific and generalized pretrained models in ArcGIS for enhanced vision-language AI applications\",\"breadcrumb\":{\"@id\":\"https:\/\/www.esri.com\/arcgis-blog\/products\/arcgis\/geoai\/task-and-generalized-pretrained-models#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/www.esri.com\/arcgis-blog\/products\/arcgis\/geoai\/task-and-generalized-pretrained-models\"]}]},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/www.esri.com\/arcgis-blog\/products\/arcgis\/geoai\/task-and-generalized-pretrained-models#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\/\/www.esri.com\/arcgis-blog\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Pretrained Models in ArcGIS: Comparing Task-Specific and Generalized Vision-Language models\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/www.esri.com\/arcgis-blog\/#website\",\"url\":\"https:\/\/www.esri.com\/arcgis-blog\/\",\"name\":\"ArcGIS Blog\",\"description\":\"Get insider info from Esri product teams\",\"publisher\":{\"@id\":\"https:\/\/www.esri.com\/arcgis-blog\/#organization\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/www.esri.com\/arcgis-blog\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"},{\"@type\":\"Organization\",\"@id\":\"https:\/\/www.esri.com\/arcgis-blog\/#organization\",\"name\":\"Esri\",\"url\":\"https:\/\/www.esri.com\/arcgis-blog\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\/\/www.esri.com\/arcgis-blog\/#\/schema\/logo\/image\/\",\"url\":\"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2018\/04\/Esri.png\",\"contentUrl\":\"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2018\/04\/Esri.png\",\"width\":400,\"height\":400,\"caption\":\"Esri\"},\"image\":{\"@id\":\"https:\/\/www.esri.com\/arcgis-blog\/#\/schema\/logo\/image\/\"},\"sameAs\":[\"https:\/\/www.facebook.com\/esrigis\/\",\"https:\/\/x.com\/ESRI\",\"https:\/\/www.linkedin.com\/company\/5311\/\"]},{\"@type\":\"Person\",\"@id\":\"https:\/\/www.esri.com\/arcgis-blog\/#\/schema\/person\/fe6ac8773de48adda0651f5f144d2acf\",\"name\":\"Vinay Viswambharan\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\/\/www.esri.com\/arcgis-blog\/#\/schema\/person\/image\/\",\"url\":\"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2018\/10\/vin4.png\",\"contentUrl\":\"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2018\/10\/vin4.png\",\"caption\":\"Vinay Viswambharan\"},\"description\":\"Principal Product manager on the Imagery team at Esri, with a zeal for remote sensing, AI and everything imagery.\",\"sameAs\":[\"https:\/\/www.esri.com\/arcgis-blog\/author\/vinayv\/\",\"https:\/\/www.linkedin.com\/in\/vinayviswambharan\/\"],\"url\":\"https:\/\/www.esri.com\/arcgis-blog\/author\/vinayv\"}]}<\/script>\n<!-- \/ Yoast SEO Premium plugin. -->","yoast_head_json":{"title":"Task vs Generalized Vision Models in ArcGIS","description":"Explore the differences between task-specific and generalized pretrained models in ArcGIS for enhanced vision-language AI applications","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/www.esri.com\/arcgis-blog\/products\/arcgis\/geoai\/task-and-generalized-pretrained-models","og_locale":"en_US","og_type":"article","og_title":"Pretrained Models in ArcGIS: Comparing Task-Specific and Generalized Vision-Language models","og_description":"Explore the differences between task-specific and generalized pretrained models in ArcGIS for enhanced vision-language AI applications","og_url":"https:\/\/www.esri.com\/arcgis-blog\/products\/arcgis\/geoai\/task-and-generalized-pretrained-models","og_site_name":"ArcGIS Blog","article_publisher":"https:\/\/www.facebook.com\/esrigis\/","article_modified_time":"2025-05-17T01:15:47+00:00","twitter_card":"summary_large_image","twitter_site":"@ESRI","twitter_misc":{"Est. reading time":"6 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":["Article","BlogPosting"],"@id":"https:\/\/www.esri.com\/arcgis-blog\/products\/arcgis\/geoai\/task-and-generalized-pretrained-models#article","isPartOf":{"@id":"https:\/\/www.esri.com\/arcgis-blog\/products\/arcgis\/geoai\/task-and-generalized-pretrained-models"},"author":{"name":"Vinay Viswambharan","@id":"https:\/\/www.esri.com\/arcgis-blog\/#\/schema\/person\/fe6ac8773de48adda0651f5f144d2acf"},"headline":"Pretrained Models in ArcGIS: Comparing Task-Specific and Generalized Vision-Language models","datePublished":"2025-05-12T18:31:52+00:00","dateModified":"2025-05-17T01:15:47+00:00","mainEntityOfPage":{"@id":"https:\/\/www.esri.com\/arcgis-blog\/products\/arcgis\/geoai\/task-and-generalized-pretrained-models"},"wordCount":10,"commentCount":0,"publisher":{"@id":"https:\/\/www.esri.com\/arcgis-blog\/#organization"},"keywords":["arcgis geoai","deep learning","geoAI","pretrained models"],"articleSection":["AI","Imagery & Remote Sensing"],"inLanguage":"en-US","potentialAction":[{"@type":"CommentAction","name":"Comment","target":["https:\/\/www.esri.com\/arcgis-blog\/products\/arcgis\/geoai\/task-and-generalized-pretrained-models#respond"]}]},{"@type":"WebPage","@id":"https:\/\/www.esri.com\/arcgis-blog\/products\/arcgis\/geoai\/task-and-generalized-pretrained-models","url":"https:\/\/www.esri.com\/arcgis-blog\/products\/arcgis\/geoai\/task-and-generalized-pretrained-models","name":"Task vs Generalized Vision Models in ArcGIS","isPartOf":{"@id":"https:\/\/www.esri.com\/arcgis-blog\/#website"},"datePublished":"2025-05-12T18:31:52+00:00","dateModified":"2025-05-17T01:15:47+00:00","description":"Explore the differences between task-specific and generalized pretrained models in ArcGIS for enhanced vision-language AI applications","breadcrumb":{"@id":"https:\/\/www.esri.com\/arcgis-blog\/products\/arcgis\/geoai\/task-and-generalized-pretrained-models#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/www.esri.com\/arcgis-blog\/products\/arcgis\/geoai\/task-and-generalized-pretrained-models"]}]},{"@type":"BreadcrumbList","@id":"https:\/\/www.esri.com\/arcgis-blog\/products\/arcgis\/geoai\/task-and-generalized-pretrained-models#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/www.esri.com\/arcgis-blog\/"},{"@type":"ListItem","position":2,"name":"Pretrained Models in ArcGIS: Comparing Task-Specific and Generalized Vision-Language models"}]},{"@type":"WebSite","@id":"https:\/\/www.esri.com\/arcgis-blog\/#website","url":"https:\/\/www.esri.com\/arcgis-blog\/","name":"ArcGIS Blog","description":"Get insider info from Esri product teams","publisher":{"@id":"https:\/\/www.esri.com\/arcgis-blog\/#organization"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/www.esri.com\/arcgis-blog\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"},{"@type":"Organization","@id":"https:\/\/www.esri.com\/arcgis-blog\/#organization","name":"Esri","url":"https:\/\/www.esri.com\/arcgis-blog\/","logo":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/www.esri.com\/arcgis-blog\/#\/schema\/logo\/image\/","url":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2018\/04\/Esri.png","contentUrl":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2018\/04\/Esri.png","width":400,"height":400,"caption":"Esri"},"image":{"@id":"https:\/\/www.esri.com\/arcgis-blog\/#\/schema\/logo\/image\/"},"sameAs":["https:\/\/www.facebook.com\/esrigis\/","https:\/\/x.com\/ESRI","https:\/\/www.linkedin.com\/company\/5311\/"]},{"@type":"Person","@id":"https:\/\/www.esri.com\/arcgis-blog\/#\/schema\/person\/fe6ac8773de48adda0651f5f144d2acf","name":"Vinay Viswambharan","image":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/www.esri.com\/arcgis-blog\/#\/schema\/person\/image\/","url":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2018\/10\/vin4.png","contentUrl":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2018\/10\/vin4.png","caption":"Vinay Viswambharan"},"description":"Principal Product manager on the Imagery team at Esri, with a zeal for remote sensing, AI and everything imagery.","sameAs":["https:\/\/www.esri.com\/arcgis-blog\/author\/vinayv\/","https:\/\/www.linkedin.com\/in\/vinayviswambharan\/"],"url":"https:\/\/www.esri.com\/arcgis-blog\/author\/vinayv"}]}},"text_date":"May 12, 2025","author_name":"Multiple Authors","author_page":"https:\/\/www.esri.com\/arcgis-blog\/products\/arcgis\/geoai\/task-and-generalized-pretrained-models","custom_image":"https:\/\/www.esri.com\/arcgis-blog\/app\/uploads\/2023\/12\/banner.gif","primary_product":"ArcGIS","tag_data":[{"term_id":779412,"name":"arcgis geoai","slug":"arcgis-geoai","term_group":0,"term_taxonomy_id":779412,"taxonomy":"post_tag","description":"","parent":0,"count":3,"filter":"raw"},{"term_id":186132,"name":"deep learning","slug":"deep-learning","term_group":0,"term_taxonomy_id":186132,"taxonomy":"post_tag","description":"","parent":0,"count":65,"filter":"raw"},{"term_id":665211,"name":"geoAI","slug":"geoai","term_group":0,"term_taxonomy_id":665211,"taxonomy":"post_tag","description":"","parent":0,"count":36,"filter":"raw"},{"term_id":758872,"name":"pretrained models","slug":"pretrained-models","term_group":0,"term_taxonomy_id":758872,"taxonomy":"post_tag","description":"","parent":0,"count":5,"filter":"raw"}],"category_data":[{"term_id":770712,"name":"AI","slug":"geoai","term_group":0,"term_taxonomy_id":770712,"taxonomy":"category","description":"","parent":0,"count":52,"filter":"raw"},{"term_id":22931,"name":"Imagery & Remote Sensing","slug":"imagery","term_group":0,"term_taxonomy_id":22931,"taxonomy":"category","description":"","parent":0,"count":778,"filter":"raw"}],"product_data":[{"term_id":421922,"name":"ArcGIS","slug":"arcgis","term_group":0,"term_taxonomy_id":421922,"taxonomy":"product","description":"Reserved for articles that cover all of ArcGIS","parent":36981,"count":347,"filter":"raw"},{"term_id":36841,"name":"ArcGIS API for Python","slug":"api-python","term_group":0,"term_taxonomy_id":36841,"taxonomy":"product","description":"","parent":36601,"count":151,"filter":"raw"},{"term_id":36951,"name":"ArcGIS Image Server","slug":"image-server","term_group":0,"term_taxonomy_id":36951,"taxonomy":"product","description":"","parent":36571,"count":69,"filter":"raw"},{"term_id":36581,"name":"ArcGIS Living Atlas","slug":"arcgis-living-atlas","term_group":0,"term_taxonomy_id":36581,"taxonomy":"product","description":"","parent":0,"count":1178,"filter":"raw"},{"term_id":36551,"name":"ArcGIS Online","slug":"arcgis-online","term_group":0,"term_taxonomy_id":36551,"taxonomy":"product","description":"","parent":0,"count":2445,"filter":"raw"},{"term_id":36561,"name":"ArcGIS Pro","slug":"arcgis-pro","term_group":0,"term_taxonomy_id":36561,"taxonomy":"product","description":"","parent":0,"count":2081,"filter":"raw"}],"primary_product_link":"https:\/\/www.esri.com\/arcgis-blog\/?s=#&products=arcgis","_links":{"self":[{"href":"https:\/\/www.esri.com\/arcgis-blog\/wp-json\/wp\/v2\/blog\/2789822","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.esri.com\/arcgis-blog\/wp-json\/wp\/v2\/blog"}],"about":[{"href":"https:\/\/www.esri.com\/arcgis-blog\/wp-json\/wp\/v2\/types\/blog"}],"author":[{"embeddable":true,"href":"https:\/\/www.esri.com\/arcgis-blog\/wp-json\/wp\/v2\/users\/8452"}],"replies":[{"embeddable":true,"href":"https:\/\/www.esri.com\/arcgis-blog\/wp-json\/wp\/v2\/comments?post=2789822"}],"version-history":[{"count":0,"href":"https:\/\/www.esri.com\/arcgis-blog\/wp-json\/wp\/v2\/blog\/2789822\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.esri.com\/arcgis-blog\/wp-json\/wp\/v2\/media?parent=2789822"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.esri.com\/arcgis-blog\/wp-json\/wp\/v2\/categories?post=2789822"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.esri.com\/arcgis-blog\/wp-json\/wp\/v2\/tags?post=2789822"},{"taxonomy":"industry","embeddable":true,"href":"https:\/\/www.esri.com\/arcgis-blog\/wp-json\/wp\/v2\/industry?post=2789822"},{"taxonomy":"product","embeddable":true,"href":"https:\/\/www.esri.com\/arcgis-blog\/wp-json\/wp\/v2\/product?post=2789822"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}