tmp

.gitignore

@@ -153,4 +153,4 @@ version.py
 
 *.out
 
-slurm_logs
+BK_folder

README.md

@@ -1,51 +1,81 @@
-## HowTOs
+# HowTOs
 
-### ✨**Installation**
+***善用加粗、倾斜***
 
-1. Clone repo
+### 📖 **RESIN: A self-supervised framework for enhancing axial resolution of volumetric imaging data**
+> [![arXiv](https://img.shields.io/badge/arXiv-Paper-<COLOR>.svg)](https://baidu.com)<br>
+> [Author](https://github.com)
+
+Based on BasicSR.
+
+### 🚩 Updates
+* **2024.9.16**: XXX
+
+## Web Demo and API
+
+[![Replicate](https://replicate.com/cjwbw/animesr/badge)]() 
+
+## Video Demos
+
+*Coming soon*
+
+## 🔧 Dependencies and Installation
+- Python >= 3.7 (Recommend to use Anaconda or Miniconda)
+- PyTorch >= 1.7
+- Other required packages in requirements.txt
+
+### **Installation**
+
+1. Clone the repo.
 
     ```bash
     git clone https://github.com/wwbwang/sr_3dunet
     cd sr_3dunet
     ```
-2. Install dependent packages and RESIN
+2. Install dependent packages and RESIN.
 
     ```bash
     # Install dependent packages
     python==3.9.0
     pip install basicsr
     pip install -r requirements.txt
 
-    # Install sr_3dunet
+    # Install RESIN
     python setup.py develop
     ```
 
-3. Switch Branch to RESIN
+3. Switch Branch to RESIN.
     ```bash
     git checkout RESIN
     ```
 
-### ✨**Start training**
-We take the de-anisotropy pipeline with VISoR's NISSL datasets for example.
-
-#### **Axial anisotropy**
-1. Get a big ROI regin from your data, save it to h5 (or tiff).
-2. Crop the above big single data to sub-images, 128×128, for example. The training patch is usually small, and get these patch should has a rule. Note that you can set some percentile rules during selecting these sub-images, and the size of this sub-image is different from the traing patch_size, the dataloader will furter randomly crop the sub-images to GT_size×GT_size patchs for training.
-Run the script extract_subimages.py:
-    ```bash
-    # 
-    bash scripts/
-    ```
-
-
-支持neuron，nissl，dapi等
-
-### ✨**Inference**
+## ⚡ **Quick Inference**
 
 ```bash
 # inference tif files in folder
 bash inference_from_folder.sh
 
 # inference a single h5 file
 bash inference_from_h5.sh
-```
+```
+
+## 💻 Training
+
+### **Dataset Preparation**
+
+1. Get a big ROI regin from your data, the format of your data should be h5/ims (or tif/tiff).
+2. Crop the above big single image to sub-images, 128×128, for example. Note that you can apply filtering rules (using args arguments or write yourself's code) when selecting data subsets, as ensuring the cleanliness of the dataset is essential for successful training.
+    
+    The scripts is as follows:
+    ```bash
+    # Get sub-images from a single tifffile
+    python scripts/get_subdatasets_from_bigtif.py --size=128 --percentiles_lower_bound=0.75 --percentiles_upper_bound=0.99 --minmax=450 --maxmin=60000 --input_folder="path" --output_folder="path" --input_tif_name="name" --crop_size=50
+    # Get sub-images from a single h5file
+    python scripts/get_subdatasets_from_ims.py --size=128 --percentiles_lower_bound=0 --percentiles_upper_bound=0.9999 --minmax=400 --maxmin=60000 --input_folder="path" --output_folder="path" --input_tif_name="name" --x_floor=0 --x_upper=8700 --y_floor=0 --y_upper=8400 --z_floor=0 --z_upper=900 --channel=1
+    ```
+    If the direction of anisotropy is not axial but 45° orientation, add `--rotated_flag` at the end of the command. The py_script will generate 2 (or 3 if rotated_flag is True) datasets, as `frontground_datasets`, `background_datasets` (and `rotated_frontground_datatsets`). Note that the `background_datasets` is just for check.
+
+### **Start Training**
+Let's take the de-anisotropy pipeline with VISoR's NISSL datasets for example.
+
+支持neuron，nissl，dapi等