Merge branch 'master' of https://github.com/ShiqiYu/OpenGait

README.md

@@ -13,7 +13,7 @@ OpenGait is a flexible and extensible gait recognition project provided by the [
 
 
 ## Our Publications
-- [CVPR 2023] LIDAR GAIT: Benchmarking 3D Gait Recognition with Point Clouds, [*Paper*](https://arxiv.org/pdf/2211.10598), [*Dataset and Code*](https://lidargait.github.io).
+- [CVPR 2023] LidarGait: Benchmarking 3D Gait Recognition with Point Clouds, [*Paper*](https://arxiv.org/pdf/2211.10598), [*Dataset and Code(Coming Soon)*](https://lidargait.github.io).
 - [CVPR 2023] OpenGait: Revisiting Gait Recognition Toward Better Practicality, [*Paper*](https://arxiv.org/pdf/2211.06597.pdf), [*Code*](configs/gaitbase).
 - [ECCV 2022] GaitEdge: Beyond Plain End-to-end Gait Recognition for Better Practicality, [*Paper*](https://arxiv.org/pdf/2203.03972), [*Code*](configs/gaitedge/README.md).
 

configs/gaitbase/gaitbase_da_casiab.yaml

@@ -9,9 +9,8 @@ data_cfg:
 evaluator_cfg:
   enable_float16: true
   restore_ckpt_strict: true
-  restore_hint: 0
+  restore_hint: 60000
   save_name: GaitBase_DA
-  #eval_func: GREW_submission
   sampler:
     batch_shuffle: false
     batch_size: 16

configs/gaitbase/gaitbase_da_gait3d.yaml

@@ -19,7 +19,7 @@ evaluator_cfg:
     frames_all_limit: 720 # limit the number of sampled frames to prevent out of memory
   metric: euc # cos
   transform:
-    - type: BaseSilTransform
+    - type: BaseSilCuttingTransform
 
 loss_cfg:
   - loss_term_weight: 1.0
@@ -103,7 +103,7 @@ trainer_cfg:
       trf_cfg:
         - type: RandomPerspective
           prob: 0.2
-        - type: BaseSilTransform
+        - type: BaseSilCuttingTransform
         - type: RandomHorizontalFlip
           prob: 0.2
         - type: RandomRotate

configs/gaitbase/gaitbase_oumvlp.yaml

@@ -11,12 +11,10 @@ evaluator_cfg:
   restore_ckpt_strict: true
   restore_hint: 120000
   save_name: GaitBase
-  #eval_func: GREW_submission
   sampler:
     batch_shuffle: false
     batch_size: 16
-    sample_type: all_ordered # all indicates whole sequence used to test, while ordered means input sequence by its natural order; 
-Other options:   fixed_unordered
+    sample_type: all_ordered # all indicates whole sequence used to test, while ordered means input sequence by its natural order; Other options:   fixed_unordered
     frames_all_limit: 720 # limit the number of sampled frames to prevent out of memory
   metric: euc # cos
   transform:

datasets/Gait3D/README.md

@@ -11,7 +11,7 @@ python datasets/pretreatment_smpl.py --input_path 'Gait3D/3D_SMPLs' --output_pat
 
 python datasets/Gait3D/merge_two_modality.py --sils_path 'Gait3D-sils-64-64-pkl' --smpls_path 'Gait3D-smpls-pkl' --output_path 'Gait3D-merged-pkl' --link 'hard'
 ```
-**Note**: If you use the processed pickle files directly, then the size of silhouette is `64x44`, which means that the pixels on both sides of the horizontal direction can no longer be cut when transforming.
+
 ## Train
 ### Baseline model:
 `CUDA_VISIBLE_DEVICES=0,1,2,3 python -m torch.distributed.launch --nproc_per_node=4 opengait/main.py --cfgs ./configs/baseline/baseline_Gait3D.yaml --phase train`