Fatal destination path during the custom object tracking evaluation

[lakyfarky]

Search before asking

• I have searched the Yolo Tracking issues and discussions and found no similar questions.

Yolo Tracking Component

Evaluation

Bug

Hi, I followed custom tracking evaluation and step 0 is okay. I set my dataset into the corresponding structure and then this error happens:

Environment

Google colab

Minimal Reproducible Example

!python examples/val.py --yolo-model "/content/drive/MyDrive/master/el_yolov8s.pt" --benchmark MOTCUSTOM --split test --tracking-method strongsort

[mikel-brostrom]

Thx for pointing this out @lakyfarky! Check this line of code

https://github.com/mikel-brostrom/yolo_tracking/blob/7ff8de0f344dfd9a8cab1c4f2b4fa729a8ba98e8/examples/val.py#L238

as a workaround for you right now, update it to:

"--SPLIT_TO_EVAL", "test",

[lakyfarky]

Thanks, it works until the tracking step. I rewrite code as in multi-class eval to adapt script to my classes. Error is distractor class 'person_on_vehicle'.

!python examples/val.py --yolo-model "/content/drive/MyDrive/master/yolov8s_vd.pt" --benchmark MOTCUSTOM --split val --tracking-method strongsort --classes 0 1 2 3 4 5 6 7 8 9

2023-08-21 08:10:05.689 | ERROR | main:eval:256 -
Eval Config:
USE_PARALLEL : True
NUM_PARALLEL_CORES : 4
BREAK_ON_ERROR : True
RETURN_ON_ERROR : False
LOG_ON_ERROR : /content/yolo_tracking/examples/val_utils/error_log.txt
PRINT_RESULTS : True
PRINT_ONLY_COMBINED : False
PRINT_CONFIG : True
TIME_PROGRESS : True
DISPLAY_LESS_PROGRESS : False
OUTPUT_SUMMARY : True
OUTPUT_EMPTY_CLASSES : True
OUTPUT_DETAILED : True
PLOT_CURVES : True

MotChallenge2DBox Config:
PRINT_CONFIG : True
GT_FOLDER : /content/yolo_tracking/examples/val_utils/data/MOTCUSTOM/val
TRACKERS_FOLDER : /content/yolo_tracking/runs/val/exp4
OUTPUT_FOLDER : None
TRACKERS_TO_EVAL : ['mot']
CLASSES_TO_EVAL : ['pedestrian', 'people', 'bicycle', 'car', 'van', 'truck', 'tricycle', 'awning-tricycle', 'bus', 'motor']
BENCHMARK :
SPLIT_TO_EVAL : val
INPUT_AS_ZIP : False
DO_PREPROC : True
TRACKER_SUB_FOLDER :
OUTPUT_SUB_FOLDER :
TRACKER_DISPLAY_NAMES : None
SEQMAP_FOLDER : None
SEQMAP_FILE : None
SEQ_INFO : {'MOTCUSTOM-01': None}
GT_LOC_FORMAT : {gt_folder}/{seq}/gt/gt.txt
SKIP_SPLIT_FOL : True

CLEAR Config:
METRICS : ['HOTA', 'CLEAR', 'Identity']
THRESHOLD : 0.5
PRINT_CONFIG : True

Identity Config:
METRICS : ['HOTA', 'CLEAR', 'Identity']
THRESHOLD : 0.5
PRINT_CONFIG : True

Evaluating 1 tracker(s) on 1 sequence(s) for 10 class(es) on MotChallenge2DBox dataset using the following metrics: HOTA, CLEAR, Identity, Count

Evaluating mot

Tracker mot was unable to be evaluated.
'person_on_vehicle'

[mikel-brostrom]

TrackEval is not build for multi-class evaluations. I have at least never manage to get it working.

[mikel-brostrom]

Check this out: #927

[lakyfarky]

Thanks, I'll try it.