Simultaneous Detection and Segmentation

Simultaneous Detection and Segmentation. Hariharan, B., Arbeláez, P., Girshick, R., & Malik, J. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 8695 LNCS(PART 7):297-312, Springer Verlag, 7, 2014.

Paper

Simultaneous Detection and Segmentation [link]

Website doi abstract bibtex

We aim to detect all instances of a category in an image and, for each instance, mark the pixels that belong to it. We call this task Simultaneous Detection and Segmentation (SDS). Unlike classical bounding box detection, SDS requires a segmentation and not just a box. Unlike classical semantic segmentation, we require individual object instances. We build on recent work that uses convolutional neural networks to classify category-independent region proposals (R-CNN [16]), introducing a novel architecture tailored for SDS. We then use category-specific, top- down figure-ground predictions to refine our bottom-up proposals. We show a 7 point boost (16% relative) over our baselines on SDS, a 5 point boost (10% relative) over state-of-the-art on semantic segmentation, and state-of-the-art performance in object detection. Finally, we provide diagnostic tools that unpack performance and provide directions for future work.

@article{
 title = {Simultaneous Detection and Segmentation},
 type = {article},
 year = {2014},
 keywords = {convolutional networks,detection,segmentation},
 pages = {297-312},
 volume = {8695 LNCS},
 websites = {https://arxiv.org/abs/1407.1808v1},
 month = {7},
 publisher = {Springer Verlag},
 day = {7},
 id = {4bc4ffc5-5c14-32ca-9e63-30ec86536430},
 created = {2023-11-06T13:55:49.716Z},
 accessed = {2023-11-06},
 file_attached = {true},
 profile_id = {f1f70cad-e32d-3de2-a3c0-be1736cb88be},
 group_id = {5ec9cc91-a5d6-3de5-82f3-3ef3d98a89c1},
 last_modified = {2023-11-08T09:05:19.376Z},
 read = {true},
 starred = {false},
 authored = {false},
 confirmed = {false},
 hidden = {false},
 folder_uuids = {bc66e353-ef41-46d4-8108-778d5481c126,bd3c6f2e-3514-47cf-bc42-12db8b9abe45},
 private_publication = {false},
 abstract = {We aim to detect all instances of a category in an image and, for each
instance, mark the pixels that belong to it. We call this task Simultaneous
Detection and Segmentation (SDS). Unlike classical bounding box detection, SDS
requires a segmentation and not just a box. Unlike classical semantic
segmentation, we require individual object instances. We build on recent work
that uses convolutional neural networks to classify category-independent region
proposals (R-CNN [16]), introducing a novel architecture tailored for SDS. We
then use category-specific, top- down figure-ground predictions to refine our
bottom-up proposals. We show a 7 point boost (16% relative) over our baselines
on SDS, a 5 point boost (10% relative) over state-of-the-art on semantic
segmentation, and state-of-the-art performance in object detection. Finally, we
provide diagnostic tools that unpack performance and provide directions for
future work.},
 bibtype = {article},
 author = {Hariharan, Bharath and Arbeláez, Pablo and Girshick, Ross and Malik, Jitendra},
 doi = {10.1007/978-3-319-10584-0_20},
 journal = {Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)},
 number = {PART 7}
}

Downloads: 0

{"_id":"rDhCevfhfCociM4sN","bibbaseid":"hariharan-arbelez-girshick-malik-simultaneousdetectionandsegmentation-2014","downloads":0,"creationDate":"2017-03-02T15:49:35.309Z","title":"Simultaneous Detection and Segmentation","author_short":["Hariharan, B.","Arbeláez, P.","Girshick, R.","Malik, J."],"year":2014,"bibtype":"article","biburl":"https://bibbase.org/service/mendeley/bfbbf840-4c42-3914-a463-19024f50b30c","bibdata":{"title":"Simultaneous Detection and Segmentation","type":"article","year":"2014","keywords":"convolutional networks,detection,segmentation","pages":"297-312","volume":"8695 LNCS","websites":"https://arxiv.org/abs/1407.1808v1","month":"7","publisher":"Springer Verlag","day":"7","id":"4bc4ffc5-5c14-32ca-9e63-30ec86536430","created":"2023-11-06T13:55:49.716Z","accessed":"2023-11-06","file_attached":"true","profile_id":"f1f70cad-e32d-3de2-a3c0-be1736cb88be","group_id":"5ec9cc91-a5d6-3de5-82f3-3ef3d98a89c1","last_modified":"2023-11-08T09:05:19.376Z","read":"true","starred":false,"authored":false,"confirmed":false,"hidden":false,"folder_uuids":"bc66e353-ef41-46d4-8108-778d5481c126,bd3c6f2e-3514-47cf-bc42-12db8b9abe45","private_publication":false,"abstract":"We aim to detect all instances of a category in an image and, for each\ninstance, mark the pixels that belong to it. We call this task Simultaneous\nDetection and Segmentation (SDS). Unlike classical bounding box detection, SDS\nrequires a segmentation and not just a box. Unlike classical semantic\nsegmentation, we require individual object instances. We build on recent work\nthat uses convolutional neural networks to classify category-independent region\nproposals (R-CNN [16]), introducing a novel architecture tailored for SDS. We\nthen use category-specific, top- down figure-ground predictions to refine our\nbottom-up proposals. We show a 7 point boost (16% relative) over our baselines\non SDS, a 5 point boost (10% relative) over state-of-the-art on semantic\nsegmentation, and state-of-the-art performance in object detection. Finally, we\nprovide diagnostic tools that unpack performance and provide directions for\nfuture work.","bibtype":"article","author":"Hariharan, Bharath and Arbeláez, Pablo and Girshick, Ross and Malik, Jitendra","doi":"10.1007/978-3-319-10584-0_20","journal":"Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)","number":"PART 7","bibtex":"@article{\n title = {Simultaneous Detection and Segmentation},\n type = {article},\n year = {2014},\n keywords = {convolutional networks,detection,segmentation},\n pages = {297-312},\n volume = {8695 LNCS},\n websites = {https://arxiv.org/abs/1407.1808v1},\n month = {7},\n publisher = {Springer Verlag},\n day = {7},\n id = {4bc4ffc5-5c14-32ca-9e63-30ec86536430},\n created = {2023-11-06T13:55:49.716Z},\n accessed = {2023-11-06},\n file_attached = {true},\n profile_id = {f1f70cad-e32d-3de2-a3c0-be1736cb88be},\n group_id = {5ec9cc91-a5d6-3de5-82f3-3ef3d98a89c1},\n last_modified = {2023-11-08T09:05:19.376Z},\n read = {true},\n starred = {false},\n authored = {false},\n confirmed = {false},\n hidden = {false},\n folder_uuids = {bc66e353-ef41-46d4-8108-778d5481c126,bd3c6f2e-3514-47cf-bc42-12db8b9abe45},\n private_publication = {false},\n abstract = {We aim to detect all instances of a category in an image and, for each\ninstance, mark the pixels that belong to it. We call this task Simultaneous\nDetection and Segmentation (SDS). Unlike classical bounding box detection, SDS\nrequires a segmentation and not just a box. Unlike classical semantic\nsegmentation, we require individual object instances. We build on recent work\nthat uses convolutional neural networks to classify category-independent region\nproposals (R-CNN [16]), introducing a novel architecture tailored for SDS. We\nthen use category-specific, top- down figure-ground predictions to refine our\nbottom-up proposals. We show a 7 point boost (16% relative) over our baselines\non SDS, a 5 point boost (10% relative) over state-of-the-art on semantic\nsegmentation, and state-of-the-art performance in object detection. Finally, we\nprovide diagnostic tools that unpack performance and provide directions for\nfuture work.},\n bibtype = {article},\n author = {Hariharan, Bharath and Arbeláez, Pablo and Girshick, Ross and Malik, Jitendra},\n doi = {10.1007/978-3-319-10584-0_20},\n journal = {Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)},\n number = {PART 7}\n}","author_short":["Hariharan, B.","Arbeláez, P.","Girshick, R.","Malik, J."],"urls":{"Paper":"https://bibbase.org/service/mendeley/bfbbf840-4c42-3914-a463-19024f50b30c/file/23a3d532-667a-bd9b-2b75-d3e737e268d6/full_text.pdf.pdf","Website":"https://arxiv.org/abs/1407.1808v1"},"biburl":"https://bibbase.org/service/mendeley/bfbbf840-4c42-3914-a463-19024f50b30c","bibbaseid":"hariharan-arbelez-girshick-malik-simultaneousdetectionandsegmentation-2014","role":"author","keyword":["convolutional networks","detection","segmentation"],"metadata":{"authorlinks":{"hariharan, b":"https://leelabcnbc.github.io/"}},"downloads":0},"search_terms":["simultaneous","detection","segmentation","hariharan","arbeláez","girshick","malik"],"keywords":["convolutional networks","detection","segmentation"],"authorIDs":["58ce9f3c2fe957a83f00000e","o5Pcovw8KYgPCuYrq"],"dataSources":["bzxc3uBcwMv3h47xE","CxzfuXz38fxepy3pc","KZFesWbmGy4yc4ZLC","ya2CyA73rpZseyrZ8","2252seNhipfTmjEBQ"]}