Papers › Template-Based Automatic Search of Compact Semantic Segmentation Architectures

Template-Based Automatic Search of Compact Semantic Segmentation Architectures

4 Apr 2019arXiv:1904.02365archive 2025-07-28

Vladimir Nekrasov, Chunhua Shen, Ian Reid

Automatic search of neural architectures for various vision and natural language tasks is becoming a prominent tool as it allows to discover high-performing structures on any dataset of interest. Nevertheless, on more difficult domains, such as dense per-pixel classification, current automatic approaches are limited in their scope - due to their strong reliance on existing image classifiers they tend to search only for a handful of additional layers with discovered architectures still containing a large number of parameters. In contrast, in this work we propose a novel solution able to find light-weight and accurate segmentation architectures starting from only few blocks of a pre-trained classification network. To this end, we progressively build up a methodology that relies on templates of sets of operations, predicts which template and how many times should be applied at each step, while also generating the connectivity structure and downsampling factors. All these decisions are being made by a recurrent neural network that is rewarded based on the score of the emitted architecture on the holdout set and trained using reinforcement learning. One discovered architecture achieves 63.2% mean IoU on CamVid and 67.8% on CityScapes having only 270K parameters. Pre-trained models and the search code are available at https://github.com/DrSleep/nas-segm-pytorch.

PaperPDFCode

Code

drsleep/nas-segm-pytorch officialmentioned in papermentioned on GitHubpytorchNOASSERTION report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

General ClassificationHoldout SetReal-Time Semantic SegmentationReinforcement LearningSemantic Segmentation

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Real-Time Semantic Segmentation Cityscapes test Template-Based-NAS-arch1 Frame (fps) 10 #33 of 39 Archive leaderboard report
Real-Time Semantic Segmentation Cityscapes test Template-Based-NAS-arch1 Time (ms) 97 #33 of 39 Archive leaderboard report
Real-Time Semantic Segmentation Cityscapes test Template-Based-NAS-arch1 mIoU 67.8% #33 of 39 Archive leaderboard report
Real-Time Semantic Segmentation Cityscapes test Template-Based-NAS-arch0 Frame (fps) 19 #34 of 39 Archive leaderboard report
Real-Time Semantic Segmentation Cityscapes test Template-Based-NAS-arch0 Time (ms) 52 #34 of 39 Archive leaderboard report
Real-Time Semantic Segmentation Cityscapes test Template-Based-NAS-arch0 mIoU 67.7% #34 of 39 Archive leaderboard report
Semantic Segmentation CamVid Template-Based NAS-arch0 (480x360 inputs) Mean IoU 63.9% #16 of 21 Archive leaderboard report
Semantic Segmentation CamVid Template-Based NAS-arch1 (480x360 inputs) Mean IoU 63.2% #18 of 21 Archive leaderboard report
Semantic Segmentation Cityscapes test Template-Based NAS-arch1 Mean IoU (class) 67.8% #87 of 105 Archive leaderboard report
Semantic Segmentation Cityscapes test Template-Based NAS-arch0 Mean IoU (class) 67.7% #88 of 105 Archive leaderboard report
Semantic Segmentation Cityscapes val Template-Based NAS-arch1 mIoU 69.5% #87 of 99 Archive leaderboard report
Semantic Segmentation Cityscapes val Template-Based NAS-arch0 mIoU 68.1% #90 of 99 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections