caffe_cls_demo

Introduction

train image classfication network by caffe

This repo is organized as follows:

caffe_cls_demo/
    ├── data
        ├── imgs
        └── labels
    ├── README.md
    ├── caffe_pred.py
    ├── lmdb_io.py
    ├── deploy.prototxt
    └── solver.prototxt

Requirements

1. caffe==1.0.0
2. python==3.6.8
3. numpy
4. opencv==3.4.1
5. Python packages might missing. pls fix it according to the error message.

Installation, Prepare data, Training

Installation

1. Clone the caffe_cls_caffe repository.

git clone https://github.com/vicwer/caffe_cls_caffe.git

2. Create data, lmdb and models directory.

Prepare data

data should be organized as follows:

data/
    |->labels/train.txt
    |->labels/test.txt
    |->imgs/*.png

1. Download dataset

2. Generate img_list.txt formatted as "img_path label"

3. Generate train.txt and test.txt formatted as "img_path label"

4. Generate lmdb:

python3 lmdb_io.py

Train

caffe train --solver=deploy.prototxt -gpu 0

Test:

python3 caffe_pred.py

GOOD LUCK...