ayh015 commited on
Commit
2c46c4d
·
1 Parent(s): 3a1265d

Restore README

Browse files
Files changed (1) hide show
  1. README.md +91 -3
README.md CHANGED
@@ -1,3 +1,91 @@
1
- ---
2
- license: mit
3
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # LLM auto annotation for HICO-DET dataset (Pose from [Halpe](https://github.com/Fang-Haoshu/Halpe-FullBody), Part State from [HAKE](https://github.com/DirtyHarryLYL/HAKE)).
2
+
3
+ ## Environment
4
+ The code is developed using python 3.11.11 on Ubuntu 21.xx with torch==2.6.0+cu124,
5
+ transformers==4.57.3 (with Qwen3 series)
6
+
7
+ ## Quick start
8
+ ### Installation
9
+ 1. Install required packges and dependencies.
10
+ 2. Clone this repo, and we'll call the directory that you cloned as ${ROOT}.
11
+ 3. Creat necessary directories:
12
+ ```
13
+ mkdir outputs
14
+ mkdir model_weights
15
+ ```
16
+ 4. Download LLM's weights into model_weights from hugging face.
17
+
18
+
19
+ ### Prepare Dataset
20
+ 5. Download [dataset]().
21
+ 6. Organize dataset, your directory tree of dataset should look like this (there maybe extra files.):
22
+ ```
23
+ {DATA_ROOT}
24
+ |-- Annotation
25
+ | |--hico-det-instance-level
26
+ | | |--hico-det-training-set-instance-level.json
27
+ | `--hico-fullbody-pose
28
+ | |--halpe_train_v1.json
29
+ | `--halpe_val_v1.json
30
+ |── Configs
31
+ | |--hico_hoi_list.txt
32
+ | `--Part_State_76.txt
33
+ |── Images
34
+ | |--images
35
+ | |--test2015
36
+ | | |--HICO_test2015_00000001.jpg
37
+ | | |--HICO_test2015_00000002.jpg
38
+ | | ...
39
+ | `--train2015
40
+ | |--HICO_train2015_00000001.jpg
41
+ | |--HICO_train2015_00000002.jpg
42
+ | ...
43
+ `── Logic_Rules
44
+ |--gather_rule.pkl
45
+ `--read_rules.py
46
+ ```
47
+
48
+ ### Start annotation
49
+ #### Modify the data_path, model_path, output_dir='outputs' by your configuration in "{ROOT}/scripts/annotate.sh".
50
+ ```
51
+ IDX={YOUR_GPU_IDS}
52
+ export PYTHONPATH=$PYTHONPATH:./
53
+
54
+ data_path={DATA_ROOT}
55
+ model_path={ROOT}/model_weights/{YOUR_MODEL_NAME}
56
+ output_dir={ROOT}/outputs
57
+
58
+ if [ -d ${output_dir} ];then
59
+ echo "dir already exists"
60
+ else
61
+ mkdir ${output_dir}
62
+ fi
63
+
64
+ CUDA_VISIBLE_DEVICES=$IDX OMP_NUM_THREADS=1 torchrun --nnodes=1 --nproc_per_node={NUM_YOUR_GPUs} --master_port=25005 \
65
+ tools/annotate.py \
66
+ --model-path ${model_path} \
67
+ --data-path ${data_path} \
68
+ --output-dir ${output_dir} \
69
+ ```
70
+ #### Start auto-annotation
71
+ ```
72
+ bash scripts/annotate.sh
73
+ ```
74
+
75
+ ## Annotation format
76
+ A list of dict that contains the following keys:
77
+ ```
78
+ {
79
+ 'file_name': 'HICO_train2015_00009511.jpg',
80
+ 'image_id': 0,
81
+ 'keypoints': a 51-elements list (17x3 keypoints with x, y, v),
82
+ 'vis': a 51-elements list (17 keypionts, each has 3 visiblity flags),
83
+ 'instance_id':0,
84
+ 'action_labels': [{'human_part': part_id, 'partstate': state_id}, ...],
85
+ 'height': 640,
86
+ 'width': 480,
87
+ 'human_bbox': [126, 258, 150, 305],
88
+ 'object_bbox': [128, 276, 144, 313],
89
+ 'description': "The person is riding a bicycle, supported by visible evidence of their body interacting with the bike.\n\n- The right hand is holding the right handlebar.\n- The left hand is holding the left handlebar.\n- The right hip is positioned over the seat, indicating the person is sitting on the bicycle.\n- The right foot is on the right pedal.\n- The left foot is on the left pedal."
90
+ }
91
+ ```