Research map / Detection and segmentation / Open-set and few-shot detection
Unsupervised Learning of Important Objects From First-Person Videos
Gedas Bertasius, Hyun Soo Park, Stella X. Yu, Jianbo Shi
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use one visual and one spatial pathway instead of two visual (ablation) | VSN · FP-AO-RGBD | max F-score: 0.402 -> 0.421 | Improved |
| use one visual and one spatial pathway instead of two spatial (ablation) | VSN · FP-AO-RGBD | max F-score: 0.400 -> 0.421 | Improved |
| use location prior (0.6W,0.75H) instead of center prior (ablation) | VSN · FP-AO-RGBD | max F-score: 0.22 -> 0.407 | Improved |
| estimate the location prior with a hand detector instead of guessing (ablation) | VSN · FP-AO-RGBD | max F-score: (0.542W,0.713H) vs (0.6W,0.75H) | No clear effect |
| train on much more unlabeled data than the 4247 labeled-size set (ablation) | VSN · FP-AO-RGBD | max F-score: 0.316 -> 0.421 | Improved |
| use hand-closest detected objects as predictions instead of VSN (ablation) | hand-object baseline · FP-AO-RGBD | max F-score: 0.259 vs 0.421 | Hurt |
| average VSN predictions with EgoNet | EgoNet · FP-AO-RGBD | max F-score: 0.396 -> 0.455 | Improved |
| average VSN predictions with EgoNet | EgoNet · GTEA Gaze+ | max F-score: 0.536 -> 0.588 | Improved |
| add VSN training on unlabeled data versus supervised EgoNet | VSN · GTEA Gaze+ | max F-score: 0.536 -> 0.482 | Hurt |
| train the network with VSN rather than plain VGG FCN pretrained initialization | VGG FCN · FP-AO-RGBD | max F-score: 0.166 -> 0.421 | Improved |
Closest papers by meaning
- Benchmarking Unsupervised Object Representations for Video Sequences (JMLR 2021 · similarity 0.87)
- Unsupervised Object Localization: Observing the Background To Discover Objects (CVPR 2023 · similarity 0.86)
- Spatially Invariant Unsupervised Object Detection with Convolutional Neural Networks (AAAI 2019 · similarity 0.86)
- Unsupervised Learning From Video to Detect Foreground Objects in Single Images (ICCV 2017 · similarity 0.86)
- Is ImageNet worth 1 video? Learning strong image encoders from 1 long unlabelled video (ICLR 2024 · similarity 0.84)
- Learning to Segment Referred Objects from Narrated Egocentric Videos (CVPR 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
