zhangxinkang/tensorflow-slim-picture-classification

Use tensorflow-slim for image classification. Use the food-101 dataset to train your own image classification model.

python train_image_classifierV1.py \
    --train_dir=weights \
    --dataset_dir=train \
    --num_samples=3320 \
    --num_classes=101 \
    --labels_to_names_path=labels.txt \
    --model_name=inception_v3 \
    --checkpoint_path=inception_v3/inception_v3.ckpt \
    --checkpoint_exclude_scopes=InceptionV3/Logits,InceptionV3/AuxLogits \
    --trainable_scopes=InceptionV3/Logits,InceptionV3/AuxLogits
--model_name 为模型名称你可以换用slim提供的其他模型
如果不需要预训练请使用一下指令
python train_image_classifierV1.py \
    --train_dir=weights \
    --dataset_dir=train \
    --num_samples=3320 \
    --num_classes=101 \
    --labels_to_names_path=labels.txt \
    --model_name=inception_v3 \

python eval_image_classifierV1.py \
    --checkpoint_path=weights \
    --eval_dir=eval \
    --dataset_dir=val \
    --num_samples=350 \
    --num_classes=101 \
    --model_name=inception_v3
--dataset_dir为验证（测试）tfrecord文件的存放位置
val.py为多张照片测试脚本drawing里面存放要测试的照片
 python val.py \
    --checkpoint_path=/home/kyz/tensorflow/main/180504/weights/model.ckpt-200000 \
    --test_dir=/home/kyz/tensorflow/main/180429/slim/drawing 


tensorboard --logdir=weights