test_produce_maps.py

import os
os.environ["CUDA_VISIBLE_DEVICES"] = "0"
import torch
import torch.nn.functional as F
import sys
import torch.nn as nn
import numpy as np
import argparse
import cv2
from Code.lib.model import TMSOD
from Code.utils.data import test_dataset
import time

parser = argparse.ArgumentParser()
parser.add_argument('--testsize', type=int, default=384, help='testing size')
parser.add_argument('--test_path', type=str, default='', help='test dataset path')
opt = parser.parse_args()

dataset_path = opt.test_path

# load the model
model = TMSOD()
model.cuda()

#set checkpoint path
model.load_state_dict({k.replace('module.', ''):v for k,v in torch.load('', map_location='cuda:0').items()})
model.eval()

# test
# ['VT2000_unalign','VT5000-Test_unalign', 'VT1000_unalign', 'VT821_unalign'] for unaligned RGBT model
# ['STERE1000', 'SIP', 'NJUD', 'NLPR', 'DUT-RGBD', 'SSD'] for RGBD model
# ['VT5000/Test', 'VT1000', 'VT821'] for aligned RGBT model
# ['ECSSD', 'HKU-IS', 'OMRON', 'PASCAL-S', 'DUTS-TE'] for RGB model
test_datasets = ['']

#test_datasets = ['NLPR', 'STERE1000', 'SIP', 'NJUD', 'NLPR', 'DUT-RGBD', 'SSD']#['STERE1000', 'SIP', 'NJUD', 'NLPR', 'DUT-RGBD', 'SSD']
for dataset in test_datasets:
    save_path = './test_maps/' + dataset + '/'
    if not os.path.exists(save_path):
        os.makedirs(save_path)

    image_root = dataset_path + dataset + '/RGB/'
    gt_root = dataset_path + dataset + '/GT/'
    depth_root = dataset_path + dataset + '/T/'
    test_loader = test_dataset(image_root, gt_root, depth_root, opt.testsize)
    img_num = len(test_loader)
    time_s = time.time()
    for i in range(test_loader.size):
        image, gt, depth, name, image_for_post = test_loader.load_data()

        gt = np.asarray(gt, np.float32)
        gt /= (gt.max() + 1e-8)
        image = image.cuda()
        depth = depth.cuda()
        pre_res = model(image, depth)
        res = pre_res
        res = F.upsample(res, size=gt.shape, mode='bilinear', align_corners=False)
        res = res.sigmoid().data.cpu().numpy().squeeze()
        res = (res - res.min()) / (res.max() - res.min() + 1e-8)

        print('save img to: ', save_path + name)
        cv2.imwrite(save_path + name, res * 255)
    time_e = time.time()
    print('speed: %f FPS' % (img_num / (time_e - time_s)))
    print('Test Done!')