Just like how SSD inference exports the output kitti label directory for all the predictions with the different model score thresholds, I would like to do the same with frcnn inference.
But, on the documentation, I do not see any arguments on frcnn - like SSD has for enabling the directory to output KITTI labels.
The tlt-infer can generate kitti format labels.
See jupyter notebook as below.
8. Visualize inferences
In this section, we run the tlt-infer tool to generate inferences on the trained models.
Running inference for detection on n images
Please go to $USER_EXPERIMENT_DIR/data/faster_rcnn/inference_results_imgs_retrain to see the visualizatons.
!tlt-infer faster_rcnn -e $SPECS_DIR/default_spec_resnet18_retrain_spec.txt
The
tlt-infertool produces two outputs.
- Overlain images in
$USER_EXPERIMENT_DIR/data/faster_rcnn/inference_results_imgs_retrain- Frame by frame bbox labels in kitti format located in
$USER_EXPERIMENT_DIR/data/faster_rcnn/inference_dump_labels_retrain