Please guide what am I doing wrong?

python run_infer.py --nr_types=5 --type_info_path=/home/hovernet/mydrive/HoverNet/hover_net/type_info.json --model_path=logs/01/net_epoch=50.tar --model_mode=original tile --input_dir=/home/hovernet/mydrive/HoverNet/hover_net/dataset/CoNSeP/Test/Images --output_dir=/home/hovernet/mydrive/HoverNet/hover_net/logs/infer_output
I am using this command for inference

You are not doing anything wrong. The current saving format for .mat is infact quite different compared to the format expected by the old stat code. Rerun inference should now have format as expected by metric code. Still, for consep you need to uncomment

You are not doing anything wrong. The current saving format for .mat is infact quite different compared to the format expected by the old stat code. Rerun inference should now have format as expected by metric code. Still, for consep you need to uncomment

hover_net/compute_stats.py

Lines 62 to 64 in a0f80c7

	# * for converting the GT type in CoNSeP
	# true_inst_type[(true_inst_type == 3) | (true_inst_type == 4)] = 3
	# true_inst_type[(true_inst_type == 5) | (true_inst_type == 6) | (true_inst_type == 7)] = 4

Still same error, compute_stats.py is not working for type mode.

I failed to get compute_stats output using this command:

python compute_stats.py --mode=type --pred_dir=/home/hovernet/mydrive/HoverNet/hover_net/logs/infer_output/mat --true_dir=/home/hovernet/mydrive/HoverNet/hover_net/dataset/CoNSeP/Test/Labels

Traceback (most recent call last):
File "compute_stats.py", line 248, in
run_nuclei_type_stat(args.pred_dir, args.true_dir)
File "compute_stats.py", line 104, in run_nuclei_type_stat
paired_all = np.concatenate(paired_all, axis=0)
File "<array_function internals>", line 6, in concatenate
ValueError: need at least one array to concatenate

Apparently, the code failed to get the correct .mat files in the predictions folder. After opening the prediction .mat file, I only got "inst_map" and inst_type. I wonder there is no "inst_centroid" in the prediction folder got through inference.

Same issue as earlier, maybe the order of dictionary keys in prediction .mat files causing error. Shouldn't the order of inst_uid and inst_type be the same as GT ?

@vqdang Thank you so much, now it's working. I just had to add "/" at the end of the path to make it run.