ruiyum / mvc

MVP-N: A Dataset and Benchmark for Real-World Multi-View Object Classification (NeurIPS 2022)

Setup

Step 1: Get repository

https://github.com/RuiyuM/MVC.git
cd MVC

Dataset Preparation make sure the dataset name is: modelnet40v2png_ori4

Step 1: Download the ModelNet40 dataset (20 view setting) and extract it to the current folder:

wget https://data.airc.aist.go.jp/kanezaki.asako/data/modelnet40v2png_ori4.tar
tar -xvf modelnet40v2png_ori4.tar

Step 2: Place data.zip in this repository Step 3: Unzip data.zip

unzip data.zip

feature aggregation QUERIES_STRATEGY = ['dissimilarity_sampling', 'uncertainty', 'random', 'patch_based_selection']

CUDA_VISIBLE_DEVICES=3 nohup python main_multi_view.py -MV_FLAG=TRAIN -MV_TYPE=MVT -QUERIES_STRATEGY=patch_based_selection -DATA_SET=MVP_N -Optimizer_selection=SGD -Query_batch=2 -initial_images=5 > patch_2_MVP_N_160_validation.log &

CUDA_VISIBLE_DEVICES=3 nohup python main_multi_view.py -MV_FLAG=TRAIN -MV_TYPE=MVT -QUERIES_STRATEGY=reverse_patch_based_selection -DATA_SET=MVP_N -Optimizer_selection=SGD -Query_batch=2 -initial_images=5 > reverse_2_MVP_N_160_validation.log &

CUDA_VISIBLE_DEVICES=3 nohup python main_multi_view.py -MV_FLAG=TRAIN -MV_TYPE=MVT -QUERIES_STRATEGY=random -DATA_SET=MVP_N -Optimizer_selection=SGD -Query_batch=2 -initial_images=5 > random_2_MVP_N_160_validation.log &