RuntimeError: invalid argument 5: k not in range for dimension about centernet HOT 11 OPEN

@WuChannn 'num_dets' should be less than or equal to 'top_k' * 'top_k'

from centernet.

@Duankaiwen Hello, kaiwen. I change top_k to 6, and num_dets in the follwing code is 8, however the num of predicted center points is 0.
def _decode( tl_heat, br_heat, tl_tag, br_tag, tl_regr, br_regr, ct_heat, ct_regr, K=100, kernel=1, ae_threshold=1, num_dets=8 ):

and the full log is:

cfg_file: config/CenterNet-52.json [6/1884]
loading all datasets...
split: test
loading from cache file: cache/ks_test.pkl
loading annotations into memory...
Done (t=1.75s)
creating index...
index created!
system config...
{'batch_size': 3,
'cache_dir': 'cache',
'chunk_sizes': [3],
'config_dir': 'config',
'data_dir': './data',
'data_rng': <mtrand.RandomState object at 0x7fec9843c8b8>,
'dataset': 'KS',
'decay_rate': 10,
'display': 50,
'learning_rate': 0.00025,
'max_iter': 480000,
'nnet_rng': <mtrand.RandomState object at 0x7fec9843c900>,
'opt_algo': 'adam',
'prefetch_size': 6,
'pretrain': None,
'result_dir': 'results',
'sampling_function': 'kp_detection',
'snapshot': 5000,
'snapshot_name': 'CenterNet-52',
'stepsize': 450000,
'test_split': 'test',
'train_split': 'train',
'val_iter': 100,
'val_split': 'test',
'weight_decay': False,
'weight_decay_rate': 1e-05,
'weight_decay_type': 'l2'}
db config...
{'ae_threshold': 0.5,
'border': 128,
'categories': 6,
'data_aug': True,
'gaussian_bump': True,
'gaussian_iou': 0.7,
'gaussian_radius': -1,
'input_size': [511, 511],
'kp_categories': 1,
'lighting': True,
'max_per_image': 100,
'merge_bbox': False,
'nms_algorithm': 'exp_soft_nms',
'nms_kernel': 3,
'nms_threshold': 0.5,
'output_sizes': [[128, 128]],
'rand_color': True,
'rand_crop': True,
'rand_pushes': False,
'rand_samples': False,
'rand_scale_max': 1.4,
'rand_scale_min': 0.6,
'rand_scale_step': 0.1,
'rand_scales': array([0.6, 0.7, 0.8, 0.9, 1. , 1.1, 1.2, 1.3]),
'special_crop': False,
'test_scales': [0.5],
'top_k': 6,
'weight_exp': 8}
loading parameters at iteration: 5000
building neural network...
module_file: models.CenterNet-52
total parameters: 104787098
loading parameters...
loading model from cache/nnet/CenterNet-52/CenterNet-52_5000.pkl
locating kps: 0%| | 0/2772 [00:00<?, ?it/s]/root/data/anaconda2/envs/CenterNet/lib/python3.6/site-packages/torch/nn/modules/upsampling.py:122: UserW
arning: nn.Upsampling is deprecated. Use nn.functional.interpolate instead.
warnings.warn("nn.Upsampling is deprecated. Use nn.functional.interpolate instead.")

Traceback (most recent call last):
File "test.py", line 94, in
test(testing_db, args.split, args.testiter, args.debug, args.suffix)
File "test.py", line 61, in test
testing(db, nnet, result_dir, debug=debug)
File "/root/data/ks/code/CenterNet_duan/test/ks.py", line 321, in testing
return globals()[system_configs.sampling_function](db, nnet, result_dir, debug=debug)
File "/root/data/ks/code/CenterNet_duan/test/ks.py", line 152, in kp_detection
center_points = np.concatenate(center_points, axis=1)
ValueError: need at least one array to concatenate

from centernet.

@WuChannn Comment out line 147 in test/coco.py

from centernet.

@Duankaiwen Thanks a lot. Hope to help others.

from centernet.

@Duankaiwen Hello, kaiwen.

When I try to test with my own images with my trained model, I always get nothing in the saved images and results.json.
I print out the result of decode_func() function in line129 in test/coco.py, and I get:

dets: [[[ 3.2349873e+01  6.6429825e+01  1.5968109e+02  2.2367534e+02
    4.6260887e-01  4.7123554e-01  4.5398217e-01  3.0000000e+00]
  [ 3.2349873e+01  6.6429825e+01  1.3143454e+02  2.2368355e+02
    4.1034997e-01  4.7123554e-01  3.4946439e-01  3.0000000e+00]
  [ 3.2349873e+01  6.6429825e+01  1.5440904e+02  2.2367509e+02
    4.0282422e-01  4.7123554e-01  3.3441293e-01  3.0000000e+00]
  [ 8.8468765e+01  6.0515965e+01  1.5968109e+02  2.2367534e+02
   -1.0000000e+00  4.9183124e-01  4.5398217e-01  3.0000000e+00]
  [ 8.8468765e+01  6.0515965e+01  5.4405136e+01  2.2369923e+02
   -1.0000000e+00  4.9183124e-01  3.7416297e-01  3.0000000e+00]
  [ 8.8468765e+01  6.0515965e+01  1.3143454e+02  2.2368355e+02
   -1.0000000e+00  4.9183124e-01  3.4946439e-01  3.0000000e+00]
  [ 8.8468765e+01  6.0515965e+01  1.5968109e+02  2.2367534e+02
   -1.0000000e+00  4.9183124e-01  4.1429698e-01  3.0000000e+00]
  [ 8.8468765e+01  6.0515965e+01  4.5412842e+01  2.2369164e+02
   -1.0000000e+00  4.9183124e-01  3.3460695e-01  3.0000000e+00]]

 [[ 1.8584467e-03  7.0376465e+01  7.1446548e+01  2.2367661e+02
    4.0953285e-01  3.7950289e-01  4.3956283e-01  3.0000000e+00]
  [ 6.4745412e+00  5.5450581e+01  1.5968164e+02  2.2367366e+02
    3.9658430e-01  3.5217959e-01  4.4098902e-01  3.0000000e+00]
  [ 6.4745412e+00  5.5450581e+01  7.1446548e+01  2.2367661e+02
    3.9587122e-01  3.5217959e-01  4.3956283e-01  3.0000000e+00]
  [ 2.7386642e+01  7.0421577e+01  1.5968164e+02  2.2367366e+02
    3.9371496e-01  3.4644091e-01  4.4098902e-01  3.0000000e+00]
  [ 2.7386642e+01  7.0421577e+01  7.1446548e+01  2.2367661e+02
    3.9300185e-01  3.4644091e-01  4.3956283e-01  3.0000000e+00]
  [ 1.8584467e-03  7.0376465e+01  6.2447205e+01  2.2368639e+02
    3.8339013e-01  3.7950289e-01  3.8727733e-01  3.0000000e+00]
  [ 1.8584467e-03  7.0376465e+01  8.8428604e+01  2.2368767e+02
    3.8202363e-01  3.7950289e-01  3.8454437e-01  3.0000000e+00]
  [ 1.8584467e-03  7.0376465e+01  1.2541402e+02  2.2366917e+02
    3.8006711e-01  3.7950289e-01  3.8063130e-01  3.0000000e+00]]]
center: [[[1.2947668e+02 1.1746306e+02 3.0000000e+00 1.6965330e-01]
  [1.2948651e+02 1.2747198e+02 3.0000000e+00 1.6731572e-01]
  [1.2948158e+02 1.2247352e+02 3.0000000e+00 1.5981433e-01]
  [1.2949446e+02 1.4548326e+02 3.0000000e+00 1.5187018e-01]
  [1.2950827e+02 1.0249621e+02 3.0000000e+00 1.5074971e-01]
  [1.2948874e+02 1.6947191e+02 3.0000000e+00 1.4942038e-01]]

 [[1.3146968e+02 1.2945517e+02 3.0000000e+00 1.7543328e-01]
  [1.3248714e+02 1.4547208e+02 3.0000000e+00 1.7366149e-01]
  [1.3247891e+02 1.3347208e+02 3.0000000e+00 1.7068261e-01]
  [1.3147018e+02 1.0245686e+02 3.0000000e+00 1.6493557e-01]
  [1.3248189e+02 1.1647341e+02 3.0000000e+00 1.6448633e-01]
  [1.3147867e+02 9.6464920e+01 3.0000000e+00 1.5871526e-01]]]

could you please give a brief explanation for each element in the vector? and I would like to know why the last elem in dets and the third elem in center are always 3.

I also print out detections before valid_ind = detections[:,4]> -1, I got the fifth elem always be -1, so the valid_detections should be [] after valid_ind = detections[:,4]> -1, then I get noting in the result and there is an IndexError in the Traceback.

detections: [[ 1.22197296e+02 2.34422729e+02 6.24000000e+02 8.32000000e+02 -1.00000000e+00 4.71235543e-01 4.53982174e-01 3.00000000e+00] [ 1.22197296e+02 2.34422729e+02 5.17916687e+02 8.32000000e+02 -1.00000000e+00 4.71235543e-01 3.49464387e-01 3.00000000e+00] [ 1.22197296e+02 2.34422729e+02 6.09671082e+02 8.32000000e+02 -1.00000000e+00 4.71235543e-01 3.34412932e-01 3.00000000e+00] [ 3.46322113e+02 2.10793686e+02 6.24000000e+02 8.32000000e+02 -1.00000000e+00 4.91831243e-01 4.53982174e-01 3.00000000e+00] [ 3.46322113e+02 2.10793686e+02 2.10280502e+02 8.32000000e+02 -1.00000000e+00 4.91831243e-01 3.74162972e-01 3.00000000e+00] [ 3.46322113e+02 2.10793686e+02 5.17916687e+02 8.32000000e+02 -1.00000000e+00 4.91831243e-01 3.49464387e-01 3.00000000e+00] [ 3.46322113e+02 2.10793686e+02 6.24000000e+02 8.32000000e+02 -1.00000000e+00 4.91831243e-01 4.14296985e-01 3.00000000e+00] [ 3.46322113e+02 2.10793686e+02 1.74367523e+02 8.32000000e+02 -1.00000000e+00 4.91831243e-01 3.34606946e-01 3.00000000e+00] [ 3.46660339e+02 2.50191681e+02 6.24000000e+02 8.32000000e+02 -1.00000000e+00 3.79502892e-01 4.39562827e-01 3.00000000e+00] [ 0.00000000e+00 1.90554764e+02 6.06142273e+02 8.32000000e+02 -1.00000000e+00 3.52179587e-01 4.40989017e-01 3.00000000e+00] [ 3.46660339e+02 1.90554764e+02 6.06142273e+02 8.32000000e+02 -1.00000000e+00 3.52179587e-01 4.39562827e-01 3.00000000e+00] [ 0.00000000e+00 2.50371918e+02 5.22624573e+02 8.32000000e+02 -1.00000000e+00 3.46440911e-01 4.40989017e-01 3.00000000e+00] [ 3.46660339e+02 2.50371918e+02 5.22624573e+02 8.32000000e+02 -1.00000000e+00 3.46440911e-01 4.39562827e-01 3.00000000e+00] [ 3.82601471e+02 2.50191681e+02 6.24000000e+02 8.32000000e+02 -1.00000000e+00 3.79502892e-01 3.87277335e-01 3.00000000e+00] [ 2.78838257e+02 2.50191681e+02 6.24000000e+02 8.32000000e+02 -1.00000000e+00 3.79502892e-01 3.84544373e-01 3.00000000e+00] [ 1.31127762e+02 2.50191681e+02 6.24000000e+02 8.32000000e+02 -1.00000000e+00 3.79502892e-01 3.80631298e-01 3.00000000e+00]]

Traceback (most recent call last): File "test.py", line 94, in <module> test(testing_db, args.split, args.testiter, args.debug, args.suffix) File "test.py", line 61, in test testing(db, nnet, result_dir, debug=debug) File "/root/data/ks/code/CenterNet_duan/test/ks.py", line 327, in testing return globals()[system_configs.sampling_function](db, nnet, result_dir, debug=debug) File "/root/data/ks/code/CenterNet_duan/test/ks.py", line 323, in kp_detection db.evaluate(result_json, cls_ids, image_ids) File "/root/data/ks/code/CenterNet_duan/db/ks.py", line 173, in evaluate coco_dets = coco.loadRes(result_json) File "data/coco/PythonAPI/pycocotools/coco.py", line 318, in loadRes if 'caption' in anns[0]: IndexError: list index out of range

look forwars to your reply, thanks a lot.

from centernet.

The situation happened mainly due to the wrong ground truth:
(x, y, w, h), not (x1, y1, x2, y2)

from centernet.

@WuChannn
1.'top_k' maybe too small.
2. how much iter does your model train?
3.. the class number in your own dataset should start from 1.
4. modify some codes in line 48 in db/coco.py to adapt to your own dataset.

from centernet.

@WuChannn Oh，I see

from centernet.

@Duankaiwen Hello, kaiwen

Actually, I wanna know whether 'kp_detection' means key point detection? and why "kp_categories" is set to 1 in CenterNet-xx.json?

Also where to find 'db.class_name' definition in 'cat_name = db.class_name(j)' in test/coco.py?

thanks a lot

from centernet.

@WuChannn 'kp_detection' is just a function name in sample/coco.py, ‘kp_categories’ is not used, you can delete it. 'db.class_name' is defined in db/coco.py, you can step through each line of code by pdb.

from centernet.

@Duankaiwen Hello, kaiwen:

I came across a strange problem: when I set test_scales to [1], and I get nothing in the saved results.json in debug mode, and come up

Traceback (most recent call last): 
File "test.py", line 94, in <module> test(testing_db, args.split, args.testiter, args.debug, args.suffix) 
File "test.py", line 61, in test testing(db, nnet, result_dir, debug=debug) 
File "/root/data/ks/code/CenterNet_duan/test/ks.py", line 327, in testing 
return globals()[system_configs.sampling_function](db, nnet, result_dir, debug=debug) 
File "/root/data/ks/code/CenterNet_duan/test/ks.py", line 323, in kp_detection 
db.evaluate(result_json, cls_ids, image_ids) 
File "/root/data/ks/code/CenterNet_duan/db/ks.py", line 173, in evaluate 
coco_dets = coco.loadRes(result_json) 
File "data/coco/PythonAPI/pycocotools/coco.py", line 318, in loadRes 
if 'caption' in anns[0]: IndexError: list index out of range

However, when I set test_scales to [0.1], and I get something in the saved results.json in debug mode, though they are wrong result.

I can't figure out why, so I refer to your help. Looking forward to your help.

from centernet.