This repository contains the implementation for the paper "Revisiting Few Shot Object Detection with Vision-Language Models"
-
Updated
May 30, 2025 - Python
This repository contains the implementation for the paper "Revisiting Few Shot Object Detection with Vision-Language Models"
YOLO11 fine-tuned on nuImages (7 classes, 960 px): mAP50 0.691 / mAP50-95 0.435, plus an equal-budget comparison of transfer learning vs fine-tuning vs training from scratch. MLflow + TensorBoard tracking, FiftyOne visualisation.
CMPE 249 HW1: what zero-shot COCO-to-driving 2D detection results hide (YOLO26, RT-DETR, YOLOE, Cosmos 3 Edge on KITTI/nuImages, Jetson Orin Nano)
To associate your repository with the nuimages topic, visit your repo's landing page and select "manage topics."