demalenk commited on
Commit
62f4346
·
1 Parent(s): 091b851

readme updates + ego4d download updates

Browse files
.gitignore CHANGED
@@ -4,4 +4,5 @@ ego4d/frames/
4
  *.pyc
5
  __pycache__/
6
 
7
- ego4d/process_ego4d_specs.py
 
 
4
  *.pyc
5
  __pycache__/
6
 
7
+ ego4d/process_ego4d_specs.py
8
+ ego4d/ego4d_full_videos/
README.md CHANGED
@@ -15,6 +15,19 @@ configs:
15
  - "lvos/annotations/**"
16
  ---
17
 
 
 
 
 
 
 
 
 
 
 
 
 
 
18
  # Dataset card for ITTO
19
 
20
  ITTO is a challenging new benchmark suite for evaluating and diagnosing the capabilities and limitations of point tracking methods.
@@ -46,6 +59,6 @@ pip install ego4d
46
  bash ego4d/install_ego4d.sh
47
  ```
48
 
49
- ## Runnin Evaluations with ITTO
50
 
51
  We provide evaluation scripts in the [ITTO github repo](https://github.com/ilonadem/itto), which contains dataloaders and model evaluation scripts for the numbers reported in the paper.
 
15
  - "lvos/annotations/**"
16
  ---
17
 
18
+ <div align="center">
19
+ <h1> Is This Tracker On? A Benchmark Protocol for Dynamic Tracking </h1>
20
+ </div>
21
+
22
+ <div align="center">
23
+ <!-- <a href="#model">Model</a> • -->
24
+ 🌐 <a href="http://131.215.140.160:8000/">Project Website</a> |
25
+ <!-- 🏆 <a href="https://hkust-nlp.github.io/agentboard/static/leaderboard.html">Leaderboard</a> | -->
26
+ 💻 <a href="https://github.com/ilonadem/itto/tree/main">Code</a> |
27
+ 📚 <a href=" ">Paper</a>
28
+
29
+ </div>
30
+
31
  # Dataset card for ITTO
32
 
33
  ITTO is a challenging new benchmark suite for evaluating and diagnosing the capabilities and limitations of point tracking methods.
 
59
  bash ego4d/install_ego4d.sh
60
  ```
61
 
62
+ ## Running Evaluations with ITTO
63
 
64
  We provide evaluation scripts in the [ITTO github repo](https://github.com/ilonadem/itto), which contains dataloaders and model evaluation scripts for the numbers reported in the paper.
ego4d/metadata/04046863-98c0-42a8-90f9-4191013cc252.json ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "video_id": "04046863-98c0-42a8-90f9-4191013cc252",
3
+ "video_name": "04046863-98c0-42a8-90f9-4191013cc252",
4
+ "original_fps": 30.0,
5
+ "chosen_fps": 10.0,
6
+ "stride": 5,
7
+ "start_sec": 26,
8
+ "end_sec": 56,
9
+ "num_output_frames": 143,
10
+ "original_video_shape": [
11
+ 450,
12
+ 3,
13
+ 1080,
14
+ 1920
15
+ ],
16
+ "cropped_video_shape": [
17
+ 134,
18
+ 3,
19
+ 1080,
20
+ 1920
21
+ ],
22
+ "offset": 1
23
+ }
ego4d/metadata/127392a3-036a-4af6-9db9-ca00141229db.json ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "video_id": "127392a3-036a-4af6-9db9-ca00141229db",
3
+ "video_name": "127392a3-036a-4af6-9db9-ca00141229db",
4
+ "original_fps": 30.0,
5
+ "chosen_fps": 10.0,
6
+ "stride": 5,
7
+ "start_sec": 26,
8
+ "end_sec": 56,
9
+ "num_output_frames": 143,
10
+ "original_video_shape": [
11
+ 450,
12
+ 3,
13
+ 1080,
14
+ 1920
15
+ ],
16
+ "cropped_video_shape": [
17
+ 134,
18
+ 3,
19
+ 1080,
20
+ 1920
21
+ ],
22
+ "offset": 1
23
+ }
ego4d/metadata/366b71ad-3f35-4704-b6ef-e793c7e73ac6.json ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "video_id": "366b71ad-3f35-4704-b6ef-e793c7e73ac6",
3
+ "video_name": "366b71ad-3f35-4704-b6ef-e793c7e73ac6",
4
+ "original_fps": 30.0,
5
+ "chosen_fps": 10.0,
6
+ "stride": 5,
7
+ "start_sec": 26,
8
+ "end_sec": 56,
9
+ "num_output_frames": 143,
10
+ "original_video_shape": [
11
+ 450,
12
+ 3,
13
+ 1080,
14
+ 1920
15
+ ],
16
+ "cropped_video_shape": [
17
+ 134,
18
+ 3,
19
+ 1080,
20
+ 1920
21
+ ],
22
+ "offset": 1
23
+ }
ego4d/metadata/426eb9d9-d4cf-4acc-9efb-979f61a5be91.json ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "video_id": "426eb9d9-d4cf-4acc-9efb-979f61a5be91",
3
+ "video_name": "426eb9d9-d4cf-4acc-9efb-979f61a5be91",
4
+ "original_fps": 30.0,
5
+ "chosen_fps": 10.0,
6
+ "stride": 5,
7
+ "start_sec": 26,
8
+ "end_sec": 56,
9
+ "num_output_frames": 143,
10
+ "original_video_shape": [
11
+ 450,
12
+ 3,
13
+ 1080,
14
+ 1920
15
+ ],
16
+ "cropped_video_shape": [
17
+ 134,
18
+ 3,
19
+ 1080,
20
+ 1920
21
+ ],
22
+ "offset": 1
23
+ }
ego4d/metadata/4675859e-620c-493a-b1e7-27c347074783.json ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "video_id": "4675859e-620c-493a-b1e7-27c347074783",
3
+ "video_name": "4675859e-620c-493a-b1e7-27c347074783",
4
+ "original_fps": 30.0,
5
+ "chosen_fps": 10.0,
6
+ "stride": 5,
7
+ "start_sec": 26,
8
+ "end_sec": 56,
9
+ "num_output_frames": 143,
10
+ "original_video_shape": [
11
+ 450,
12
+ 3,
13
+ 1080,
14
+ 1920
15
+ ],
16
+ "cropped_video_shape": [
17
+ 134,
18
+ 3,
19
+ 1080,
20
+ 1920
21
+ ],
22
+ "offset": 1
23
+ }
ego4d/metadata/b8b4fc74-e036-4731-ae03-4bfab7fd47b9.json ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "video_id": "b8b4fc74-e036-4731-ae03-4bfab7fd47b9",
3
+ "video_name": "b8b4fc74-e036-4731-ae03-4bfab7fd47b9",
4
+ "original_fps": 30.0,
5
+ "chosen_fps": 10.0,
6
+ "stride": 5,
7
+ "start_sec": 35,
8
+ "end_sec": 65,
9
+ "num_output_frames": 143,
10
+ "original_video_shape": [
11
+ 450,
12
+ 3,
13
+ 1080,
14
+ 1920
15
+ ],
16
+ "cropped_video_shape": [
17
+ 134,
18
+ 3,
19
+ 1080,
20
+ 1920
21
+ ],
22
+ "offset": 1
23
+ }
ego4d/metadata/be462bcf-13ad-446a-8dfb-248cb2e3417a.json ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "video_id": "be462bcf-13ad-446a-8dfb-248cb2e3417a",
3
+ "video_name": "be462bcf-13ad-446a-8dfb-248cb2e3417a",
4
+ "original_fps": 30.0,
5
+ "chosen_fps": 10.0,
6
+ "stride": 5,
7
+ "start_sec": 26,
8
+ "end_sec": 56,
9
+ "num_output_frames": 143,
10
+ "original_video_shape": [
11
+ 450,
12
+ 3,
13
+ 1080,
14
+ 1920
15
+ ],
16
+ "cropped_video_shape": [
17
+ 134,
18
+ 3,
19
+ 1080,
20
+ 1920
21
+ ],
22
+ "offset": 0
23
+ }
ego4d/metadata/f42ae9c2-d43e-45af-8adb-02d0b16b2ef7.json ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "video_id": "f42ae9c2-d43e-45af-8adb-02d0b16b2ef7",
3
+ "video_name": "f42ae9c2-d43e-45af-8adb-02d0b16b2ef7",
4
+ "original_fps": 30.0,
5
+ "chosen_fps": 10.0,
6
+ "stride": 5,
7
+ "start_sec": 26,
8
+ "end_sec": 56,
9
+ "num_output_frames": 143,
10
+ "original_video_shape": [
11
+ 450,
12
+ 3,
13
+ 1080,
14
+ 1920
15
+ ],
16
+ "cropped_video_shape": [
17
+ 134,
18
+ 3,
19
+ 1080,
20
+ 1920
21
+ ],
22
+ "offset": 1
23
+ }
ego4d/metadata/fa6386e1-82ba-4c67-b217-e11640546582.json ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "video_id": "fa6386e1-82ba-4c67-b217-e11640546582",
3
+ "video_name": "fa6386e1-82ba-4c67-b217-e11640546582",
4
+ "original_fps": 30.0,
5
+ "chosen_fps": 10.0,
6
+ "stride": 5,
7
+ "start_sec": 26,
8
+ "end_sec": 56,
9
+ "num_output_frames": 143,
10
+ "original_video_shape": [
11
+ 450,
12
+ 3,
13
+ 1080,
14
+ 1920
15
+ ],
16
+ "cropped_video_shape": [
17
+ 134,
18
+ 3,
19
+ 1080,
20
+ 1920
21
+ ],
22
+ "offset": 1
23
+ }
ego4d/process_ego4d_videos.py CHANGED
@@ -15,8 +15,10 @@ def md5(path):
15
  h.update(chunk)
16
  return h.hexdigest()
17
 
 
18
  metadata_path = "metadata"
19
- json_list = os.listdir(metadata_path)
 
20
 
21
  for json_file in json_list:
22
  with open(os.path.join("metadata", json_file), "r") as f:
@@ -32,7 +34,7 @@ for json_file in json_list:
32
  stride = metadata["stride"]
33
 
34
  # save paths
35
- video_path = f"ego4d_videos/v2/full_scale/{video_id}.mp4"
36
  save_dir = f"frames/{video_name}"
37
 
38
  print("\n PROCESSING VIDEO ", video_name)
@@ -57,7 +59,7 @@ for json_file in json_list:
57
  print(f"Saved {len(cropped_vid)} frames to {save_dir}")
58
 
59
  # # TODO(ilona) - remove this later
60
- # # checks that current dataset aligns with the files that we have !!
61
  # for frame_number in [f"{i:06d}" for i in range(0, num_output_frames)]:
62
  # paths = [
63
  # f"/data/ilona/datasets/ego4d/ego4d_chosen_videos/{video_name}/rgb_frames/{frame_number}.jpg",
@@ -100,4 +102,3 @@ for json_file in json_list:
100
  # print("stride: ", stride)
101
  # print("mean abs error: ", mae)
102
  # print("psnr: ", psnr)
103
-
 
15
  h.update(chunk)
16
  return h.hexdigest()
17
 
18
+
19
  metadata_path = "metadata"
20
+ # json_list = os.listdir(metadata_path)
21
+ json_list = ["426eb9d9-d4cf-4acc-9efb-979f61a5be91.json"]
22
 
23
  for json_file in json_list:
24
  with open(os.path.join("metadata", json_file), "r") as f:
 
34
  stride = metadata["stride"]
35
 
36
  # save paths
37
+ video_path = f"ego4d_full_videos/v2/full_scale/{video_id}.mp4"
38
  save_dir = f"frames/{video_name}"
39
 
40
  print("\n PROCESSING VIDEO ", video_name)
 
59
  print(f"Saved {len(cropped_vid)} frames to {save_dir}")
60
 
61
  # # TODO(ilona) - remove this later
62
+ # # checks that current dataset aligns with the files that we have !!
63
  # for frame_number in [f"{i:06d}" for i in range(0, num_output_frames)]:
64
  # paths = [
65
  # f"/data/ilona/datasets/ego4d/ego4d_chosen_videos/{video_name}/rgb_frames/{frame_number}.jpg",
 
102
  # print("stride: ", stride)
103
  # print("mean abs error: ", mae)
104
  # print("psnr: ", psnr)
 
ego4d/uids_test.txt ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ be462bcf-13ad-446a-8dfb-248cb2e3417a
2
+ 426eb9d9-d4cf-4acc-9efb-979f61a5be91
3
+ fa6386e1-82ba-4c67-b217-e11640546582
4
+ 366b71ad-3f35-4704-b6ef-e793c7e73ac6
5
+ f42ae9c2-d43e-45af-8adb-02d0b16b2ef7
6
+ 04046863-98c0-42a8-90f9-4191013cc252
7
+ 127392a3-036a-4af6-9db9-ca00141229db
8
+ 4675859e-620c-493a-b1e7-27c347074783
9
+ b8b4fc74-e036-4731-ae03-4bfab7fd47b9