Efficient change detection in traffic camera images using confidence and geometry filtering
CoSA: Correlation-Guided Change A ttention with Learnable Residual Gating for Remote Sensing Change Detection
Computer Vision and Pattern Recognition
Summary
It can be very expensive and difficult to label images from traffic cameras that are fixed in place. The authors studied how to improve models that were originally trained on street-level images so they work better on images taken from elevated CCTV cameras. They combined a small number of labeled CCTV images with many unlabeled ones, using confidence scores and geometric information to select the best examples for training. Their work shows that careful filtering helps improve the performance of these models, but also highlights the importance of properly evaluating the results with real human-labeled data.
remote sensingchange detectionCCTV imagerypixel-level annotationU-NetDeepLabV3pseudo-labelingintersection over unionconfidence filtering
Authors
Abdirashid Omar, Jonghyuk Park
Abstract
Pixel-level annotation of fixed traffic-camera imagery is expensive, while crosswalk models trained from street-level imagery face a substantial viewpoint and appearance shift when applied to elevated CCTV. We investigate a data-efficient target-domain pipeline using 241 manually annotated CCTV images and 5,926 unlabeled CCTV frames. A source-domain experiment trains a 31.0M-parameter custom U-Net on 3,300 first-person-view (FPV) images and obtains 93.05% IoU on its 330-image FPV test split. This result is a source baseline, not transferred performance: the released CCTV notebook instantiates a 42.0M-parameter DeepLabV3-ResNet50 from torchvision weights, and no compatible mapping from the U-Net checkpoint is implemented. Training on 201 manual CCTV images and selecting on 40 held-out manual masks yields 88.91% IoU. The model then predicts all unlabeled frames; image-level certainty and a largest-component area prior rank the candidates, and the top 1,000 attain mean certainty 0.976 and mean combined score 0.988. A repository audit shows that the reported second-stage 98.52% IoU was measured on a 150-image split containing only teacher-generated pseudo-masks. Because of a directory-layout mismatch, the executed combined-data loader found zero manual samples and split 1,000 pseudo-labeled samples into 850 training and 150 evaluation samples. We therefore report 98.52% as internal pseudo-label agreement rather than human-ground-truth accuracy. The defensible target-domain result is 88.91% IoU on the 40 manual validation images. Batch-one FP32 inference at 512 x 512 requires 12.98 ms, corresponding to 77.03 FPS, on an NVIDIA RTX A6000 48 GB GPU. These findings support the practicality of confidence-and-geometry filtering while also showing why pseudo-label evaluation must remain isolated from the labels used for self-training.