A novel few-shot object detection framework for multi-scene driving based on contrastive proposal encoding
Yalei Dong, Jing Xiao, Fengchen Wei, Jinhao Liang, Jinhao Liang, Jinhao Liang · 2025
Read A novel few-shot object detection framework for multi-scene driving based on contrastive proposal encoding by Yalei Dong, Jing Xiao, Fengchen Wei, Jinhao Liang, Jinhao Liang, Jinhao Liang (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-12752960.txt.
Description
Yalei Dong and co-authors adapt object detection for driving scenes with very little training data in a 2025 paper. A detector must spot cars and people in both daylight photos and night infrared images. With few labeled examples, models tend to overfit and fail when the scene type changes.
Their design builds on a known few-shot method and adds parts that blend local and wide context at several scales. A cosine-based classifier, which compares angles between features, helps keep each class tight. The authors say theirs is the first few-shot test across both night infrared and day visible driving scenes. On two public data sets, the method beat rivals in accuracy and in handling new scenes. Future work aims to trim the model's size.
Catalog record
| CREATOR | Yalei Dong, Jing Xiao, Fengchen Wei, Jinhao Liang, Jinhao Liang, Jinhao Liang |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 68 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |