# Labelme - [Labelme tutorial](https://hackmd.io/@agbug/ByjZArc9h) - [Labelme github](https://github.com/wkentaro/labelme) ## Preparation ### Prepare Folder 1. [Download ICALAB_val_image](https://drive.google.com/drive/folders/1XAJ9Pcwi1-pNGjYvxFnvrbDn38dt1zP4?usp=drive_link) ![](https://hackmd.io/_uploads/rkJl2dsc3.png) 3. Clone [labelme_tutorial](https://github.com/patrick0115/labelme_tutorial) or download ZIP ``` git clone git@github.com:patrick0115/labelme_tutorial.git ``` - Folder structure ```bash Anypath ├── ICALAB_val_image │ ├── JPEGImages │ ├── SegmentationClass │ ├── json │ ├── SegmentationClassPNG │ ├── SegmentationClassVisualization │ ├── color_coding_semantic_segmentation_classes - Sheet1.csv │ ├── class_names.txt │ ├── images │ │ │── 0000.jpg │ │ │── 0001.jpg │ │ │── 0002.jpg │ │ │ ....... ├── labelme_tutorial │ ├── json │ ├── output │ ├── label.txt ``` ![](https://hackmd.io/_uploads/HyAq4K55n.png) ![](https://hackmd.io/_uploads/r1RD6wc53.png) ### Install labelme ```bash pip install labelme ``` ### Open labelme ```bash labelme ``` ![](https://hackmd.io/_uploads/HyJ7RIq9n.png) ## Start label 1. Open Dir Choose images folder with raw images. ![](https://hackmd.io/_uploads/B1ZqR8q93.png) 2. Label - 描繪 ![](https://hackmd.io/_uploads/Syz1eP9qh.png) - 編輯圖案 可以移動複製,調整點位 ![](https://hackmd.io/_uploads/H1-gePcch.png) ![](https://hackmd.io/_uploads/HJoGePq52.png) - 修改類別 ![](https://hackmd.io/_uploads/B18Hxwq53.png) - 畫面色調調整 若圖片不清楚，可以調整圖案亮度 ![](https://hackmd.io/_uploads/rysTlv5c2.png) - 儲存在 labelme_tutorial/json ![](https://hackmd.io/_uploads/ByJArYc52.png) ## Start Semantic Segmentation ```bash python .\labelme2voc.py ``` - 會在output資料夾產生一下檔案 ![](https://hackmd.io/_uploads/Skw3195c3.png) - 上傳到對應資料夾(包含json檔) ![](https://hackmd.io/_uploads/BJhAjOs93.png)