图像检索-基于多个全局描述符组合实现的图像检索算法-附项目源码-优质项目实战.zip

# Combination of Multiple Global Descriptors for Image Retrieval <div align="center"> <img src="figures/architecture.png"> </div> ## Prerequisite * Python 2.7 or above * MXNet-1.4.0 or above * Numpy and tqdm ## Usage ### Download dataset ```console $ bash download.sh cub200 ``` ### Extract pre-trained model ```console $ tar zxvf ./checkpoints/CGD.CUB200.C_concat_MG.ResNet50v.dim1536.tar.gz -C ./checkpoints/ ``` ### Test ```console $ python test.py usage: test.py [-h] [--image-width IMAGE_WIDTH] [--image-height IMAGE_HEIGHT] [--batch-size BATCH_SIZE] [--num-workers NUM_WORKERS] [--recallk RECALLK] [--data-dir DATA_DIR] [--train-txt TRAIN_TXT] [--test-txt TEST_TXT] [--bbox-txt BBOX_TXT] --pretrained-model PRETRAINED_MODEL [--gpu GPU] ``` ```console $ python test.py --pretrained-model=checkpoints/CGD.CUB200.C_concat_MG.ResNet50v.dim1536 ... R@ 1: 0.7681 R@ 2: 0.8484 R@ 4: 0.9060 R@ 8: 0.9433 ```