Test Conditions for Trisoup in TMC13v12

[VMohinder]

In TMC13v12, trisoup has only r01 to r04 rates. There are no configuration files for r05 and r06 anymore.

Just to make the scripts run, I added the following function:

def set_tmc13_rate_limit(mpeg_mode):
    if mpeg_mode['label'] == 'G-PCC trisoup':
        return 4
    return 6

And then, when needed:

rSup = set_tmc13_rate_limit(mpeg_mode)
for rate in rates[0:rSup]:
          Code

Not very elegant, but solved the problem for the G-PCC experiments (python mp_run.py ev_experiment.yml)

Running experiments using the models (python ev_run_experiment.py ev_experiment.yml --num_parallel 8) still has some issues regarding trisoup rates.

python ev_run_experiment.py ev_experiment.yml --num_parallel 8 2021-03-23 12:44:27.065 INFO ev_run_experiment - : Starting our method's experiments 0%| | 0/136 [00:01<?, ?it/s] Traceback (most recent call last): File "ev_run_experiment.py", line 91, in parallel_process(run_experiment, params, args.num_parallel) File "/media/bizon/DATA01_1TB/MachineLearning/PointCloudCompression/pcc_geo_cnn_v2/src/utils/parallel_process.py", line 41, in parallel_process raise RuntimeError(f'{" ".join([shlex.quote(x) for x in p.args])} returned with code {p.returncode}\n{logs}') RuntimeError: python ev_experiment.py --output_dir /media/bizon/DATA01_1TB/MachineLearning/PointCloudCompression/pcc_geo_cnn_v2/pcc_geo_cnn_v2/soldier_vox10_0690/c2/1.00e-05 --model_dir /media/bizon/DATA01_1TB/MachineLearning/PointCloudCompression/pcc_geo_cnn_v2/pcc_geo_cnn_v2/models/c2/1.00e-05 --model_config c2 --opt_metrics d1_mse d2_mse --max_deltas inf --pc_name soldier_vox10_0690 --pcerror_path /media/bizon/DATA02_4TB/MPEG/mpeg-pcc-dmetric/test/pc_error_d --pcerror_cfg_path /media/bizon/DATA02_4TB/MPEG/TMC13/mpeg-pcc-tmc13/cfg/trisoup-predlift/lossy-geom-lossy-attrs/soldier_vox10_0690/r06/pcerror.cfg --input_pc /media/bizon/DATA01_1TB/MachineLearning/PointCloudCompression/pcc_geo_cnn_v2/data/Static_Objects_and_Scenes/People/soldier_vox10_0690/soldier_vox10_0690.ply --input_norm /media/bizon/DATA01_1TB/MachineLearning/PointCloudCompression/pcc_geo_cnn_v2/data/Static_Objects_and_Scenes/People/soldier_vox10_0690/soldier_vox10_0690_n.ply --fixed_threshold returned with code 1 Traceback (most recent call last): File "ev_experiment.py", line 194, in args.fixed_threshold, args.no_merge_coding, args.num_parallel, args.no_stream_redirection) File "ev_experiment.py", line 55, in run_experiment assert_exists(f) File "/media/bizon/DATA01_1TB/MachineLearning/PointCloudCompression/pcc_geo_cnn_v2/src/utils/experiment.py", line 8, in assert_exists assert os.path.exists(filepath), f'{filepath} not found' AssertionError: /media/bizon/DATA02_4TB/MPEG/TMC13/mpeg-pcc-tmc13/cfg/trisoup-predlift/lossy-geom-lossy-attrs/soldier_vox10_0690/r06/pcerror.cfg not found

Probably this line of code in ev_run_experiment.py must be updated:

pcerror_cfg_path = f'{MPEG_TMC13_DIR}/cfg/{pcerror_mpeg_mode}/{cfg_name}/r06/pcerror.cfg'

[mauriceqch]

Thank you!

This line

pcerror_cfg_path = f'{MPEG_TMC13_DIR}/cfg/{pcerror_mpeg_mode}/{cfg_name}/r06/pcerror.cfg

can probably be replaced by

pcerror_cfg_path = f'{MPEG_TMC13_DIR}/cfg/{pcerror_mpeg_mode}/{cfg_name}/r01/pcerror.cfg'

as the pcerror configurations seem to be the same for the rates of a single point cloud.

I'll keep this in mind for future updates.