This paper tries to address limitation of manually crafted anchor box for object detection, which is not good at handling crowed objects, occluded objects and objects with very larger/smaller aspect ratio.
\n\nModern CNN-based object detectors assign anchors for ground-truth objects under the restriction of object-anchor Intersection-over-Unit (IoU). In this study, we propose a learning-to-match approach to break IoU restriction, allowing objects to match anchors in a flexible manner. Our approach, referred to as FreeAnchor, updates hand-crafted anchor assignment to “free” anchor matching by formulating detector training as a maximum likelihood estimation (MLE) procedure. FreeAnchor targets at learning features which best explain a class of objects in terms of both classification and localization. FreeAnchor is implemented by optimizing detection customized likelihood and can be fused with CNN-based detectors in a plug-and-play manner. Experiments on MS-COCO demonstrate that FreeAnchor consistently outperforms their counterparts with significant margins.
\n\n\n\n", "messages": [ { "id": "680872", "postDate": "11/25/2019 10:55:50", "content": "This paper tries to address limitation of manually crafted anchor box for object detection, which is not good at handling crowed objects, occluded objects and objects with very larger/smaller aspect ratio.
\n\nModern CNN-based object detectors assign anchors for ground-truth objects under the restriction of object-anchor Intersection-over-Unit (IoU). In this study, we propose a learning-to-match approach to break IoU restriction, allowing objects to match anchors in a flexible manner. Our approach, referred to as FreeAnchor, updates hand-crafted anchor assignment to “free” anchor matching by formulating detector training as a maximum likelihood estimation (MLE) procedure. FreeAnchor targets at learning features which best explain a class of objects in terms of both classification and localization. FreeAnchor is implemented by optimizing detection customized likelihood and can be fused with CNN-based detectors in a plug-and-play manner. Experiments on MS-COCO demonstrate that FreeAnchor consistently outperforms their counterparts with significant margins.
\n\n\n\n", "rawMarkdown": "This paper tries to address limitation of manually crafted anchor box for object detection, which is not good at handling crowed objects, occluded objects and objects with very larger/smaller aspect ratio.\n\n**Modern CNN-based object detectors assign anchors for ground-truth objects under the restriction of object-anchor Intersection-over-Unit (IoU). In this study, we propose a learning-to-match approach to break IoU restriction, allowing objects to match anchors in a flexible manner. Our approach, referred to as FreeAnchor, updates hand-crafted anchor assignment to “free” anchor matching by formulating detector training as a maximum likelihood estimation (MLE) procedure. FreeAnchor targets at learning features which best explain a class of objects in terms of both classification and localization. FreeAnchor is implemented by optimizing detection customized likelihood and can be fused with CNN-based detectors in a plug-and-play manner. Experiments on MS-COCO demonstrate that FreeAnchor consistently outperforms their counterparts with significant margins.**\n\n**[Paper link ](https://github.com/mobassir94/ML-DL-Research-paper-Collection/blob/master/3D%20Deep%20Learning/FreeAnchor.pdf)**\n\n**[Code link ](https://github.com/zhangxiaosong18/FreeAnchor)**", "votes": null }, { "id": "681430", "postDate": "11/26/2019 05:21:14", "content": "Nice Share!! @mobassir
", "rawMarkdown": "Nice Share!! @mobassir", "votes": null }, { "id": "681435", "postDate": "11/26/2019 05:30:02", "content": "Thanks mate
", "rawMarkdown": "Thanks mate", "votes": null } ], "comments": [ { "id": 681430, "author_name": "veeralakrishna", "author_url": "", "post_date": "11/26/2019 05:21:14", "content": "Nice Share!! @mobassir
", "votes": null, "replies": [ { "id": 681435, "author_name": "mobassir", "author_url": "", "post_date": "11/26/2019 05:30:02", "content": "Thanks mate
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "680872": "This paper tries to address limitation of manually crafted anchor box for object detection, which is not good at handling crowed objects, occluded objects and objects with very larger/smaller aspect ratio.\n\n**Modern CNN-based object detectors assign anchors for ground-truth objects under the restriction of object-anchor Intersection-over-Unit (IoU). In this study, we propose a learning-to-match approach to break IoU restriction, allowing objects to match anchors in a flexible manner. Our approach, referred to as FreeAnchor, updates hand-crafted anchor assignment to “free” anchor matching by formulating detector training as a maximum likelihood estimation (MLE) procedure. FreeAnchor targets at learning features which best explain a class of objects in terms of both classification and localization. FreeAnchor is implemented by optimizing detection customized likelihood and can be fused with CNN-based detectors in a plug-and-play manner. Experiments on MS-COCO demonstrate that FreeAnchor consistently outperforms their counterparts with significant margins.**\n\n**[Paper link ](https://github.com/mobassir94/ML-DL-Research-paper-Collection/blob/master/3D%20Deep%20Learning/FreeAnchor.pdf)**\n\n**[Code link ](https://github.com/zhangxiaosong18/FreeAnchor)**", "681430": "Nice Share!! @mobassir", "681435": "Thanks mate" }, "source": "meta" }