PerceptionLM: Open-Access Data and Models for Detailed Visual Understanding
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , , , , , , , , , , , , , , , , , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866909702974078976 |
|---|---|
| author | Cho, Jang Hyun Madotto, Andrea Mavroudi, Effrosyni Afouras, Triantafyllos Nagarajan, Tushar Maaz, Muhammad Song, Yale Ma, Tengyu Hu, Shuming Jain, Suyog Martin, Miguel Wang, Huiyu Rasheed, Hanoona Sun, Peize Huang, Po-Yao Bolya, Daniel Ravi, Nikhila Jain, Shashank Stark, Tammy Moon, Shane Damavandi, Babak Lee, Vivian Westbury, Andrew Khan, Salman Krähenbühl, Philipp Dollár, Piotr Torresani, Lorenzo Grauman, Kristen Feichtenhofer, Christoph |
| author_facet | Cho, Jang Hyun Madotto, Andrea Mavroudi, Effrosyni Afouras, Triantafyllos Nagarajan, Tushar Maaz, Muhammad Song, Yale Ma, Tengyu Hu, Shuming Jain, Suyog Martin, Miguel Wang, Huiyu Rasheed, Hanoona Sun, Peize Huang, Po-Yao Bolya, Daniel Ravi, Nikhila Jain, Shashank Stark, Tammy Moon, Shane Damavandi, Babak Lee, Vivian Westbury, Andrew Khan, Salman Krähenbühl, Philipp Dollár, Piotr Torresani, Lorenzo Grauman, Kristen Feichtenhofer, Christoph |
| contents | Vision-language models are integral to computer vision research, yet many high-performing models remain closed-source, obscuring their data, design and training recipe. The research community has responded by using distillation from black-box models to label training data, achieving strong benchmark results, at the cost of measurable scientific progress. However, without knowing the details of the teacher model and its data sources, scientific progress remains difficult to measure. In this paper, we study building a Perception Language Model (PLM) in a fully open and reproducible framework for transparent research in image and video understanding. We analyze standard training pipelines without distillation from proprietary models and explore large-scale synthetic data to identify critical data gaps, particularly in detailed video understanding. To bridge these gaps, we release 2.8M human-labeled instances of fine-grained video question-answer pairs and spatio-temporally grounded video captions. Additionally, we introduce PLM-VideoBench, a suite for evaluating challenging video understanding tasks focusing on the ability to reason about "what", "where", "when", and "how" of a video. We make our work fully reproducible by providing data, training recipes, code & models. https://github.com/facebookresearch/perception_models |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2504_13180 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | PerceptionLM: Open-Access Data and Models for Detailed Visual Understanding Cho, Jang Hyun Madotto, Andrea Mavroudi, Effrosyni Afouras, Triantafyllos Nagarajan, Tushar Maaz, Muhammad Song, Yale Ma, Tengyu Hu, Shuming Jain, Suyog Martin, Miguel Wang, Huiyu Rasheed, Hanoona Sun, Peize Huang, Po-Yao Bolya, Daniel Ravi, Nikhila Jain, Shashank Stark, Tammy Moon, Shane Damavandi, Babak Lee, Vivian Westbury, Andrew Khan, Salman Krähenbühl, Philipp Dollár, Piotr Torresani, Lorenzo Grauman, Kristen Feichtenhofer, Christoph Computer Vision and Pattern Recognition Artificial Intelligence Machine Learning Vision-language models are integral to computer vision research, yet many high-performing models remain closed-source, obscuring their data, design and training recipe. The research community has responded by using distillation from black-box models to label training data, achieving strong benchmark results, at the cost of measurable scientific progress. However, without knowing the details of the teacher model and its data sources, scientific progress remains difficult to measure. In this paper, we study building a Perception Language Model (PLM) in a fully open and reproducible framework for transparent research in image and video understanding. We analyze standard training pipelines without distillation from proprietary models and explore large-scale synthetic data to identify critical data gaps, particularly in detailed video understanding. To bridge these gaps, we release 2.8M human-labeled instances of fine-grained video question-answer pairs and spatio-temporally grounded video captions. Additionally, we introduce PLM-VideoBench, a suite for evaluating challenging video understanding tasks focusing on the ability to reason about "what", "where", "when", and "how" of a video. We make our work fully reproducible by providing data, training recipes, code & models. https://github.com/facebookresearch/perception_models |
| title | PerceptionLM: Open-Access Data and Models for Detailed Visual Understanding |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence Machine Learning |
| url | https://arxiv.org/abs/2504.13180 |