i found this research paper interesting for this competition
\n\nA weakly supervised semantic segmentation (WSSS) method aims to learn a segmentation model from weak (image-level) as opposed to strong (pixel-level) labels. By avoiding the tedious pixel-level annotation process, it can exploit the unlimited supply of user-tagged images from media-sharing sites such as Flickr for large scale applications. However, these `free' tags/labels are often noisy and few existing works address the problem of learning with both weak and noisy labels. In this work, we cast the WSSS problem into a label noise reduction problem. Specifically, after segmenting each image into a set of superpixels, the weak and potentially noisy image-level labels are propagated to the superpixel level resulting in highly noisy labels; the key to semantic segmentation is thus to identify and correct the superpixel noisy labels. To this end, a novel L 1 -optimisation based sparse learning model is formulated to directly and explicitly detect noisy labels. To solve the L 1 -optimisation problem, we further develop an efficient learning algorithm by introducing an intermediate labelling variable. Extensive experiments on three benchmark datasets show that our method yields state-of-the-art results given noise-free labels, whilst significantly outperforming the existing methods when the weak labels are also noisy.
\n\nyou can get the full article directly FROM HERE
", "messages": [ { "id": "660036", "postDate": "10/28/2019 15:26:34", "content": "i found this research paper interesting for this competition
\n\nA weakly supervised semantic segmentation (WSSS) method aims to learn a segmentation model from weak (image-level) as opposed to strong (pixel-level) labels. By avoiding the tedious pixel-level annotation process, it can exploit the unlimited supply of user-tagged images from media-sharing sites such as Flickr for large scale applications. However, these `free' tags/labels are often noisy and few existing works address the problem of learning with both weak and noisy labels. In this work, we cast the WSSS problem into a label noise reduction problem. Specifically, after segmenting each image into a set of superpixels, the weak and potentially noisy image-level labels are propagated to the superpixel level resulting in highly noisy labels; the key to semantic segmentation is thus to identify and correct the superpixel noisy labels. To this end, a novel L 1 -optimisation based sparse learning model is formulated to directly and explicitly detect noisy labels. To solve the L 1 -optimisation problem, we further develop an efficient learning algorithm by introducing an intermediate labelling variable. Extensive experiments on three benchmark datasets show that our method yields state-of-the-art results given noise-free labels, whilst significantly outperforming the existing methods when the weak labels are also noisy.
\n\nyou can get the full article directly FROM HERE
", "rawMarkdown": "i found **[this research paper](https://ieeexplore.ieee.org/document/7450177)** interesting for this competition\n\n# Abstract:\nA weakly supervised semantic segmentation (WSSS) method aims to learn a segmentation model from weak (image-level) as opposed to strong (pixel-level) labels. By avoiding the tedious pixel-level annotation process, it can exploit the unlimited supply of user-tagged images from media-sharing sites such as Flickr for large scale applications. However, these `free' tags/labels are often noisy and few existing works address the problem of learning with both weak and noisy labels. In this work, we cast the WSSS problem into a label noise reduction problem. Specifically, after segmenting each image into a set of superpixels, the weak and potentially noisy image-level labels are propagated to the superpixel level resulting in highly noisy labels; the key to semantic segmentation is thus to identify and correct the superpixel noisy labels. To this end, a novel L 1 -optimisation based sparse learning model is formulated to directly and explicitly detect noisy labels. To solve the L 1 -optimisation problem, we further develop an efficient learning algorithm by introducing an intermediate labelling variable. Extensive experiments on three benchmark datasets show that our method yields state-of-the-art results given noise-free labels, whilst significantly outperforming the existing methods when the weak labels are also noisy.\n\n\nyou can get the full article directly [FROM HERE](https://github.com/mobassir94/ML-DL-Research-paper-Collection/blob/master/Deep%20Learning%20Papers/semantic%20segmentation/learning%20from%20weak%20and%20noisy%20labels.pdf)", "votes": null }, { "id": "660293", "postDate": "10/29/2019 01:02:05", "content": "Very good find , brother . I found this paper some days back, went through it , but could not understand how to implement it :)
", "rawMarkdown": "Very good find , brother . I found this paper some days back, went through it , but could not understand how to implement it :)", "votes": null }, { "id": "660438", "postDate": "10/29/2019 05:49:31", "content": "hi,i also tried the same but maximum time i get deeplabv3+ instead, so i guess deeplabv3+ is better alternative of this? i might will try deeplabv3+ @phoenix9032
", "rawMarkdown": "hi,i also tried the same but maximum time i get deeplabv3+ instead, so i guess deeplabv3+ is better alternative of this? i might will try deeplabv3+ @phoenix9032", "votes": null }, { "id": "660603", "postDate": "10/29/2019 11:15:07", "content": "Very Informative\nThanks @mobassir
", "rawMarkdown": "Very Informative\nThanks @mobassir", "votes": null }, { "id": "660817", "postDate": "10/29/2019 16:37:11", "content": "you are welcome
", "rawMarkdown": "you are welcome", "votes": null } ], "comments": [ { "id": 660293, "author_name": "phoenix9032", "author_url": "", "post_date": "10/29/2019 01:02:05", "content": "Very good find , brother . I found this paper some days back, went through it , but could not understand how to implement it :)
", "votes": null, "replies": [ { "id": 660438, "author_name": "mobassir", "author_url": "", "post_date": "10/29/2019 05:49:31", "content": "hi,i also tried the same but maximum time i get deeplabv3+ instead, so i guess deeplabv3+ is better alternative of this? i might will try deeplabv3+ @phoenix9032
", "votes": null, "replies": [] } ] }, { "id": 660603, "author_name": "veeralakrishna", "author_url": "", "post_date": "10/29/2019 11:15:07", "content": "Very Informative\nThanks @mobassir
", "votes": null, "replies": [ { "id": 660817, "author_name": "mobassir", "author_url": "", "post_date": "10/29/2019 16:37:11", "content": "you are welcome
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "660036": "i found **[this research paper](https://ieeexplore.ieee.org/document/7450177)** interesting for this competition\n\n# Abstract:\nA weakly supervised semantic segmentation (WSSS) method aims to learn a segmentation model from weak (image-level) as opposed to strong (pixel-level) labels. By avoiding the tedious pixel-level annotation process, it can exploit the unlimited supply of user-tagged images from media-sharing sites such as Flickr for large scale applications. However, these `free' tags/labels are often noisy and few existing works address the problem of learning with both weak and noisy labels. In this work, we cast the WSSS problem into a label noise reduction problem. Specifically, after segmenting each image into a set of superpixels, the weak and potentially noisy image-level labels are propagated to the superpixel level resulting in highly noisy labels; the key to semantic segmentation is thus to identify and correct the superpixel noisy labels. To this end, a novel L 1 -optimisation based sparse learning model is formulated to directly and explicitly detect noisy labels. To solve the L 1 -optimisation problem, we further develop an efficient learning algorithm by introducing an intermediate labelling variable. Extensive experiments on three benchmark datasets show that our method yields state-of-the-art results given noise-free labels, whilst significantly outperforming the existing methods when the weak labels are also noisy.\n\n\nyou can get the full article directly [FROM HERE](https://github.com/mobassir94/ML-DL-Research-paper-Collection/blob/master/Deep%20Learning%20Papers/semantic%20segmentation/learning%20from%20weak%20and%20noisy%20labels.pdf)", "660293": "Very good find , brother . I found this paper some days back, went through it , but could not understand how to implement it :)", "660438": "hi,i also tried the same but maximum time i get deeplabv3+ instead, so i guess deeplabv3+ is better alternative of this? i might will try deeplabv3+ @phoenix9032", "660603": "Very Informative\nThanks @mobassir", "660817": "you are welcome" }, "source": "meta" }