Aurelie123/DETR_Resnet50_KITTI

Model Card for DETR Facebook Resnet50

A DETR facebook Resnet50 model fine tuned on the KITTI dataset.

Model Details

Model Description

The model has been fine tuned for the task of object detection on the KITTI dataset uisng the Hugging Face dataset https://huggingface.co/datasets/nateraw/kitti .

Model Sources

Dataset : https://huggingface.co/datasets/nateraw/kitti

Original KITTI dataset source: https://www.cvlibs.net/datasets/kitti/eval_object.php

Uses

This model can be used for inference and fine-tuning on an object detection task.

Training Results

The model was trained on 10 epochs and the training took about 1 hour. The Map (precision) scores for each object class were as follows:

[Car]=0.706100

[Cyclist]=0.028800

[Dontcare]=0.387900

[Misc]=0.000800

[Pedestrian]=0.425900

[Person Sitting]=0.000000

[Tram]=0.011100

[Truck]=0.151800

[Van]=0.117700

The 'dontcare' class are areas that the model should ignore during training.

The training loss was 0.9 and the test evaluation loss was 0.85.

Model Card Authors

Aurelie