Qualitative Results

[mjjc111]

hello，author，I want to refer to the code of the "Qualitative Results" section, but I haven't found it.Is the code for the "Qualitative Results" section not made public?

[bighuang624]

Yes, it has not been included. I find one copy, but I am not sure if it is the used one (as it has been a long time, about 2 years). You may have to do some small updates if bugs happen.

'''
For qualitative analysis and other analysis
'''
import argparse
import copy
import json
import os
from itertools import product

import numpy as np
import torch
import torch.backends.cudnn as cudnn
from scipy.stats import hmean
from torch.utils.data.dataloader import DataLoader
from tqdm import tqdm
import cv2

from utils import *
from parameters import parser
from dataset import CompositionDataset
from model.model_factory import get_model
from test import predict_logits

cudnn.benchmark = True

device = "cuda" if torch.cuda.is_available() else "cpu"


if __name__ == "__main__":
    config = parser.parse_args()
    if config.yml_path:
        load_args(config.yml_path, config)

    # set the seed value
    
    print("evaluation details")
    print("----")
    print(f"dataset: {config.dataset}")

    dataset_path = config.dataset_path
    with torch.no_grad():
        print('loading test dataset')
        test_dataset = CompositionDataset(dataset_path,
                                        phase='test',
                                        split='compositional-split-natural',
                                        open_world=False)    # change this if open-world
        allattrs = test_dataset.attrs
        allobj = test_dataset.objs
        classes = [cla.replace(".", " ").lower() for cla in allobj]
        attributes = [attr.replace(".", " ").lower() for attr in allattrs]
        offset = len(attributes)
        model = get_model(config, attributes=attributes, classes=classes, offset=offset).cuda()
        model.load_state_dict(torch.load(config.load_model))
        all_logits, all_attr_gt, all_obj_gt, all_pair_gt, loss_avg = predict_logits(
            model, test_dataset, config)
        pairs_dataset = test_dataset.pairs

        print('all_logits.shape:', all_logits.shape) 

        prediction_list = []
        prediction_idx = torch.max(all_logits, dim=-1)[1]
        prediction_list = [pairs_dataset[idx.item()] for idx in prediction_idx]

        # save
        write_json(os.path.join(config.save_path, "top1_test_prediction.json"), prediction_list)
        torch.save(all_logits, os.path.join(config.save_path, "test_prediction_logits.pt"))