Update README_SBeA_tracker.md

README_SBeA_tracker.md

@@ -25,6 +25,16 @@ Case data:
 ## Run
 ### 1. Create/load project  
 SBeA_tracker is managed in a project folder. The first step is to create or load a project.  
+
+GUI:  
+ ![image](https://github.com/YNCris/SBeA_release/blob/main/demo/tracker_gui.png)   
+
+Input:  
+Create A New 'configfile.yaml'. Select a path and define a name to save your project (workspace). Or load existing 'configfile.yaml' in the first textbox.
+
+Output:  
+The workspace of your project.  
+
 Demo results:  
 ![image](https://github.com/YNCris/SBeA_release/blob/main/demo/create_project.png)  
 Notes:  
@@ -33,17 +43,18 @@ evals folder: save model evaluation results
 models folder: save deep learning models  
 configfile.yaml: the configuration of SBeA_tracker 
 
-GUI:  
- ![image](https://github.com/YNCris/SBeA_release/blob/main/demo/tracker_gui.png)   
+
+### 2. Set configurations
+The second step is to set the configurations for training.  
+
+GUI:   
+![image](https://github.com/YNCris/SBeA_release/blob/main/demo/config_gui.png)   
 
 Input:  
-Create A New 'configfile.yaml'. Select a path and define a name to save your project (workspace). Or load existing 'configfile.yaml' in the first textbox.
+Your path of social and ID data.
 
 Output:  
-The workspace of your project.
-
-### 2. Set configurations
-The second step is to set the configurations for training.  
+Changed configfile.yaml.   
 
 Your data need to be organized like:  
 Social data:  
@@ -66,17 +77,17 @@ F1: the recording serial number
 F2: the single animal name such as A1 (animal one)
 F3: the date
 
+### 3. Load data
+The third step is to load data according to configfile.yaml. If the train and ID data are not in the path in step 2, this part is going to copy them in these folders. Additionally, the raw frames, trajectories, backgrounds, and frames for labeling are extracted in this step. If the train and ID data are existing in these folders, you only need to run load train videos without filling any testbox. 
+
 GUI:   
-![image](https://github.com/YNCris/SBeA_release/blob/main/demo/config_gui.png)   
+![image](https://github.com/YNCris/SBeA_release/blob/main/demo/load_gui.png)   
 
 Input:  
-Your path of social and ID data.
+Social and ID data.
 
 Output:  
-Changed configfile.yaml. 
-
-### 3. Load data
-The third step is to load data according to configfile.yaml. If the train and ID data are not in the path in step 2, this part is going to copy them in these folders. Additionally, the raw frames, trajectories, backgrounds, and frames for labeling are extracted in this step. If the train and ID data are existing in these folders, you only need to run load train videos without filling any testbox. 
+The preprocessed data.  
 
 Demo results:  
 The raw frames in .\datasets\raw_video_images  
@@ -91,28 +102,10 @@ The trajectories in .\datasets\video_trajectories
 The manual label frames in .\datasets\manual_labels  
 ![image](https://github.com/YNCris/SBeA_release/blob/main/demo/manual_label_data.png)  
 
-GUI:   
-![image](https://github.com/YNCris/SBeA_release/blob/main/demo/load_gui.png)   
-
-Input:  
-Social and ID data.
-
-Output:  
-The preprocessed data.
 
 ### 4. Label and train
-The fourth step is to label frames and train models. In this part, you can label frames and train models in parallel, which save the running time.
-
-Demo results:  
-Label mask frames calling [labelme](https://github.com/wkentaro/labelme):  
-![image](https://github.com/YNCris/SBeA_release/blob/main/demo/label_frames.png)  
-
-Label pose frames calling [DeepLabCut](https://github.com/DeepLabCut/DeepLabCut):  
-![image](https://github.com/YNCris/SBeA_release/blob/main/demo/label_poses.png) 
+The fourth step is to label frames and train models. In this part, you can label frames and train models in parallel, which save the running time.  
 
-Training data generation based on [YOLACT++](https://github.com/dbolya/yolact):  
-![image](https://github.com/YNCris/SBeA_release/blob/main/demo/virtual_data.png) 
-
 GUI:   
 ![image](https://github.com/YNCris/SBeA_release/blob/main/demo/labeltrain_gui.png)   
 
@@ -126,9 +119,6 @@ Well-trained video instance segmentation model based on [VisTR](https://github.c
 Well-trained single animal pose estimation model. 
 Well-trained animal identification model based on [EfficientNet](https://github.com/lukemelas/EfficientNet-PyTorch). 
 
-### 4. Evaluation (Optional)
-The fourth step is to label frames and train models. In this part, you can label frames and train models in parallel, which save the running time.
-
 Demo results:  
 Label mask frames calling [labelme](https://github.com/wkentaro/labelme):  
 ![image](https://github.com/YNCris/SBeA_release/blob/main/demo/label_frames.png)  
@@ -139,6 +129,10 @@ Label pose frames calling [DeepLabCut](https://github.com/DeepLabCut/DeepLabCut)
 Training data generation based on [YOLACT++](https://github.com/dbolya/yolact):  
 ![image](https://github.com/YNCris/SBeA_release/blob/main/demo/virtual_data.png) 
 
+
+### 4. Evaluation (Optional)
+The fourth step is to label frames and train models. In this part, you can label frames and train models in parallel, which save the running time.  
+
 GUI:   
 ![image](https://github.com/YNCris/SBeA_release/blob/main/demo/labeltrain_gui.png)   
 
@@ -152,6 +146,17 @@ Well-trained video instance segmentation model based on [VisTR](https://github.c
 Well-trained single animal pose estimation model. 
 Well-trained animal identification model based on [EfficientNet](https://github.com/lukemelas/EfficientNet-PyTorch). 
 
+Demo results:  
+Label mask frames calling [labelme](https://github.com/wkentaro/labelme):  
+![image](https://github.com/YNCris/SBeA_release/blob/main/demo/label_frames.png)  
+
+Label pose frames calling [DeepLabCut](https://github.com/DeepLabCut/DeepLabCut):  
+![image](https://github.com/YNCris/SBeA_release/blob/main/demo/label_poses.png) 
+
+Training data generation based on [YOLACT++](https://github.com/dbolya/yolact):  
+![image](https://github.com/YNCris/SBeA_release/blob/main/demo/virtual_data.png) 
+
+
 ### 5. Predict
 The fifth step is to predict 3D poses with identities of new videos.