Datasets:

Modalities:
Image
Text
Formats:
webdataset
ArXiv:
Libraries:
Datasets
WebDataset
License:
evaluation/clips/clips_gt.json ADDED
The diff for this file is too large to render. See raw diff
 
evaluation/clips/eval.py ADDED
@@ -0,0 +1,59 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import argparse
2
+ import json
3
+ from pathlib import Path
4
+ import sys
5
+
6
+ def point_in_xywh(px, py, x, y, w, h):
7
+ return (x <= px < x + w) and (y <= py < y + h)
8
+
9
+ def main():
10
+
11
+ parser = argparse.ArgumentParser()
12
+
13
+ parser.add_argument(
14
+ "--predictions", type=str, required=True, help="Path to json file with predictions for each clip"
15
+ )
16
+
17
+ args = parser.parse_args()
18
+
19
+ try:
20
+ with open("clips_gt.json", 'r', encoding='utf-8') as f:
21
+ clips_gt = json.load(f)
22
+
23
+ with open(args.predictions, 'r', encoding='utf-8') as f:
24
+ preds = json.load(f)
25
+ except FileNotFoundError as e:
26
+ print(f"File not found: {e.filename}")
27
+ sys.exit(1)
28
+
29
+ except json.JSONDecodeError as e:
30
+ print(f"Invalid JSON in file: {e}")
31
+ sys.exit(1)
32
+
33
+ except Exception as e:
34
+ print(f"Unexpected error: {e}")
35
+ sys.exit(1)
36
+
37
+ total = 0
38
+ hits = 0
39
+
40
+ for clip, annots in preds.items():
41
+ if clips_gt.get(clip, None) is None:
42
+ print(f'{clip} not present in GT')
43
+ continue
44
+
45
+ for frame, point in annots.items(): #point: (x,y)
46
+ if clips_gt[clip].get(frame, None) is None:
47
+ print(f'Frame {frame} in {clip} not present in GT')
48
+ continue
49
+ total+=1
50
+ GT_box = clips_gt[clip].get(frame)
51
+ if point_in_xywh(*point, *GT_box):
52
+ hits+=1
53
+
54
+ print(f"Total: {total}, hits: {hits}")
55
+ print(f'Pointwise-Acc: {hits/total:.3f}')
56
+
57
+ if __name__ == "__main__":
58
+ main()
59
+