A new framework using faster-RCNN network has been integrated, which can be found in: object_detection/faster_rcnn.pytorch/
- Task1: Train an existing object detection network
- Task2: Read papers FasterRCNN and YOLO
- Task3: Train the network for another dataset: Pascal and Udacity
- Task4: Boost performance in both Networks
- Task5: Report and Slides
The following table shows the results obtained from the different tasks. We have used Faster RCNN network and the experiments have been done using the following datasets: Pascal and Udacity.
Network | Metric | Pascal VOC first approach | Pascal VOC Best | Udacity | |||
---|---|---|---|---|---|---|---|
Val | Test | Val | Test | Val | Test | ||
FasterRCNN | mAP | 0.66 | 0.60 | 0.92 | 0.73 | 0.81 | 0.44 |