test.py

import argparse
import os
import datetime
import logging
import time
import math
import numpy as np
from tqdm import tqdm
from collections import OrderedDict
import re
from PIL import Image

import torch
import torch.nn.functional as F

from core.configs import cfg
from core.datasets import build_dataset
from core.models import build_feature_extractor, build_classifier
from core.utils.misc import mkdir, AverageMeter, intersectionAndUnionGPU, get_color_pallete, resize
from core.utils.logger import setup_logger
from core.utils.metric_logger import MetricLogger

def strip_prefix_if_present(state_dict, prefix):
    keys = sorted(state_dict.keys())
    if not all(key.startswith(prefix) for key in keys):
        return state_dict
    stripped_state_dict = OrderedDict()
    for key, value in state_dict.items():
        stripped_state_dict[key.replace(prefix, "")] = value
    return stripped_state_dict

def inference(feature_extractor, classifier, image, label, flip=True):
    size = label.shape[-2:]
    if flip:
        image = torch.cat([image, torch.flip(image, [3])], 0)
    with torch.no_grad():
        output = classifier(feature_extractor(image))
        output = resize(
            input=output,
            size=size,
            mode='bilinear',
            align_corners=False)
    output = F.softmax(output, dim=1)
    if flip:
        output = (output[0] + output[1].flip(2)) / 2
    else:
        output = output[0]
    return output.unsqueeze(dim=0)

def multi_scale_inference(feature_extractor, classifier, image, label, scales=[0.7,1.0,1.3], flip=True):
    output = None
    size = image.shape[-2:]
    for s in scales:
        x = F.interpolate(image, size=(int(size[0]*s), int(size[1]*s)), mode='bilinear', align_corners=True)
        pred = inference(feature_extractor, classifier, x, label, flip=False)
        if output is None:
            output = pred
        else:
            output = output + pred
        if flip:
            x_flip = torch.flip(x, [3])
            pred = inference(feature_extractor, classifier, x_flip, label, flip=False)
            output = output + pred.flip(3)
    if flip:
        return output/len(scales)/2
    return output/len(scales)


def test(cfg, saveres):
    logger = logging.getLogger("TransDA.tester")
    logger.info("Start testing")
    device = torch.device(cfg.MODEL.DEVICE)

    feature_extractor = build_feature_extractor(cfg)
    feature_extractor.to(device)

    classifier,_ = build_classifier(cfg)
    classifier.to(device)
    
    if cfg.resume:
        logger.info("Loading checkpoint from {}".format(cfg.resume))
        checkpoint = torch.load(cfg.resume, map_location=torch.device('cpu'))
        feature_extractor_weights = strip_prefix_if_present(checkpoint['feature_extractor'], 'module.')
        feature_extractor.load_state_dict(feature_extractor_weights, strict = False)
        classifier_weights = strip_prefix_if_present(checkpoint['classifier'], 'module.')
        classifier.load_state_dict(classifier_weights)
        if "iteration" in checkpoint:
            iteration = checkpoint['iteration']
            print(iteration)

    feature_extractor.eval()
    classifier.eval()
    
    intersection_meter = AverageMeter()
    union_meter = AverageMeter()
    target_meter = AverageMeter()
    
    torch.cuda.empty_cache()  # TODO check if it helps
    dataset_name = cfg.DATASETS.TEST
    output_folder = '.'
    if cfg.OUTPUT_DIR:
        output_folder = os.path.join(cfg.OUTPUT_DIR, "inference", dataset_name)
        mkdir(output_folder)

    test_data = build_dataset(cfg, mode='test', is_source=False)

    test_loader = torch.utils.data.DataLoader(test_data, batch_size=cfg.TEST.BATCH_SIZE, shuffle=False, num_workers=4, pin_memory=True, sampler=None)

    
    if cfg.MODEL.NUM_CLASSES == 19:
        trainid_to_id = {
                0:7,
                1:8,
                2:11,
                3:12,
                4:13,
                5:17,
                6:19,
                7:20,
                8:21,
                9:22,
                10:23,
                11:24,
                12:25,
                13:26,
                14:27,
                15:28,
                16:31,
                17:32,
                18:33,
                255:255
            }
    else:
        trainid_to_id = {
            0:7,
            1:8,
            2:11,
            3:12,
            4:13,
            5:17,
            6:19,
            7:20,
            8:21,
            9:23,
            10:24,
            11:25,
            12:26,
            13:28,
            14:32,
            15:33,
            255:255
        }

    for batch in tqdm(test_loader):
        x, y, name = batch
        x = x.cuda(non_blocking=True)
        y = y.cuda(non_blocking=True).long()

        # if saveres:
        pred = inference(feature_extractor, classifier, x, y, flip=False)
        # else:
        # pred = multi_scale_inference(feature_extractor, classifier, x, y, flip=True)

        output = pred.max(1)[1]
        intersection, union, target = intersectionAndUnionGPU(output, y, cfg.MODEL.NUM_CLASSES, cfg.INPUT.IGNORE_LABEL)
        intersection, union, target = intersection.cpu().numpy(), union.cpu().numpy(), target.cpu().numpy()
        intersection_meter.update(intersection), union_meter.update(union), target_meter.update(target)

        accuracy = sum(intersection_meter.val) / (sum(target_meter.val) + 1e-10)

        if saveres:
            pred = pred.cpu().numpy().squeeze()
            pred = pred.argmax(0)
            mask = get_color_pallete(pred, "city")
            mask_filename = name[0] if len(name[0].split("/"))<2 else name[0].split("/")[1]
            mask.save(os.path.join(output_folder, mask_filename))
        else:
            pred = pred.cpu().numpy().squeeze()
            pred = pred.argmax(0)
            mask = 255 * np.ones(pred.shape, dtype=np.float32)
            for k, v in trainid_to_id.items():
                mask[pred == k] = v
            mask_filename = name[0] if len(name[0].split("/"))<2 else name[0].split("/")[1]
            mask = Image.fromarray(mask.astype(np.uint8), 'L')
            mask.save(os.path.join(output_folder, mask_filename))

    
    iou_class = intersection_meter.sum / (union_meter.sum + 1e-10)
    accuracy_class = intersection_meter.sum / (target_meter.sum + 1e-10)
    mIoU = np.mean(iou_class)
    mAcc = np.mean(accuracy_class)
    allAcc = sum(intersection_meter.sum) / (sum(target_meter.sum) + 1e-10)

    logger.info('Val result: mIoU/mAcc/allAcc {:.4f}/{:.4f}/{:.4f}.'.format(mIoU, mAcc, allAcc))
    for i in range(cfg.MODEL.NUM_CLASSES):
        logger.info('{} {} iou/accuracy: {:.4f}/{:.4f}.'.format(i, test_data.trainid2name[i], iou_class[i], accuracy_class[i]))


def main():
    parser = argparse.ArgumentParser(description="PyTorch Semantic Segmentation Testing")
    parser.add_argument("-cfg",
        "--config-file",
        default="",
        metavar="FILE",
        help="path to config file",
        type=str,
    )
    parser.add_argument('--saveres', action="store_true",
                        help='save the result')
    parser.add_argument(
        "opts",
        help="Modify config options using the command-line",
        default=None,
        nargs=argparse.REMAINDER,
    )

    args = parser.parse_args()

    torch.backends.cudnn.benchmark = True

    cfg.merge_from_file(args.config_file)
    cfg.merge_from_list(args.opts)
    cfg.freeze()

    output_dir = cfg.OUTPUT_DIR
    if output_dir:
        mkdir(output_dir)
    logger = setup_logger("TransDA", output_dir, 0)
    logger.info(cfg)

    logger.info("Loaded configuration file {}".format(args.config_file))
    with open(args.config_file, "r") as cf:
        config_str = "\n" + cf.read()
        logger.info(config_str)
    logger.info("Running with config:\n{}".format(cfg))
    

    test(cfg, args.saveres)


if __name__ == "__main__":
    main()