DeepLearningExamples/PyTorch/Segmentation/MaskRCNN/pytorch/tools/test_net.py

# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
# Copyright (c) 2018, NVIDIA CORPORATION. All rights reserved.
# Set up custom environment before nearly anything else is imported
# NOTE: this should be the first import (no not reorder)
from maskrcnn_benchmark.utils.env import setup_environment  # noqa F401 isort:skip

import argparse
import os

import torch
from maskrcnn_benchmark.config import cfg
from maskrcnn_benchmark.data import make_data_loader
from maskrcnn_benchmark.engine.inference import inference
from maskrcnn_benchmark.modeling.detector import build_detection_model
from maskrcnn_benchmark.utils.checkpoint import DetectronCheckpointer
from maskrcnn_benchmark.utils.collect_env import collect_env_info
from maskrcnn_benchmark.utils.comm import synchronize, get_rank, is_main_process
from maskrcnn_benchmark.utils.logger import setup_logger
from maskrcnn_benchmark.utils.miscellaneous import mkdir
from maskrcnn_benchmark.utils.logger import format_step
import dllogger

def main():
    parser = argparse.ArgumentParser(description="PyTorch Object Detection Inference")
    parser.add_argument(
        "--config-file",
        default="/workspace/object_detection/configs/e2e_mask_rcnn_R_50_FPN_1x.yaml",
        metavar="FILE",
        help="path to config file",
    )
    parser.add_argument("--local_rank", type=int, default=os.getenv('LOCAL_RANK', 0))
    parser.add_argument("--json-summary",
                        help="Out file for DLLogger",
                        default="dllogger_inference.out",
                        type=str)
    parser.add_argument(
        "--skip-eval",
        dest="skip_eval",
        help="Do not eval the predictions",
        action="store_true",
    )
    parser.add_argument(
        "--fp16",
        help="Mixed precision training",
        action="store_true",
    )
    parser.add_argument(
        "--amp",
        help="Mixed precision training",
        action="store_true",
    )
    parser.add_argument(
        "opts",
        help="Modify config options using the command-line",
        default=None,
        nargs=argparse.REMAINDER,
    )

    args = parser.parse_args()
    args.fp16 = args.fp16 or args.amp
    num_gpus = int(os.environ["WORLD_SIZE"]) if "WORLD_SIZE" in os.environ else 1
    distributed = num_gpus > 1

    if distributed:
        torch.cuda.set_device(args.local_rank)
        torch.distributed.init_process_group(
            backend="nccl", init_method="env://"
        )
        synchronize()

    cfg.merge_from_file(args.config_file)
    cfg.merge_from_list(args.opts)
    cfg.freeze()
    
    save_dir = ""
    logger = setup_logger("maskrcnn_benchmark", save_dir, get_rank())
    if is_main_process():
        dllogger.init(backends=[dllogger.JSONStreamBackend(verbosity=dllogger.Verbosity.VERBOSE,
                                filename=args.json_summary),
                                dllogger.StdOutBackend(verbosity=dllogger.Verbosity.VERBOSE, step_format=format_step)])
    else:
        dllogger.init(backends=[])


    save_dir = ""
    dllogger.log(step="PARAMETER", data={"config":cfg})
    dllogger.log(step="PARAMETER", data={"gpu_count": num_gpus})
    # dllogger.log(step="PARAMETER", data={"env_info": collect_env_info()})
    model = build_detection_model(cfg)
    model.to(cfg.MODEL.DEVICE)

    # Initialize mixed-precision
    if args.fp16:
        use_mixed_precision = True
    else:
        use_mixed_precision = cfg.DTYPE == "float16" 
    
    output_dir = cfg.OUTPUT_DIR
    checkpointer = DetectronCheckpointer(cfg, model, save_dir=output_dir)
    _ = checkpointer.load(cfg.MODEL.WEIGHT)

    iou_types = ("bbox",)
    if cfg.MODEL.MASK_ON:
        iou_types = iou_types + ("segm",)
    output_folders = [None] * len(cfg.DATASETS.TEST)
    dataset_names = cfg.DATASETS.TEST
    if cfg.OUTPUT_DIR:
        for idx, dataset_name in enumerate(dataset_names):
            output_folder = os.path.join(cfg.OUTPUT_DIR, "inference", dataset_name)
            mkdir(output_folder)
            output_folders[idx] = output_folder
    data_loaders_val = make_data_loader(cfg, is_train=False, is_distributed=distributed)

    results = []
    for output_folder, dataset_name, data_loader_val in zip(output_folders, dataset_names, data_loaders_val):
        if use_mixed_precision:
            with torch.cuda.amp.autocast():
                result = inference(
                    model,
                    data_loader_val,
                    dataset_name=dataset_name,
                    iou_types=iou_types,
                    box_only=cfg.MODEL.RPN_ONLY,
                    device=cfg.MODEL.DEVICE,
                    expected_results=cfg.TEST.EXPECTED_RESULTS,
                    expected_results_sigma_tol=cfg.TEST.EXPECTED_RESULTS_SIGMA_TOL,
                    output_folder=output_folder,
                    skip_eval=args.skip_eval,
                    dllogger=dllogger,
                )
        else:
            result = inference(
                    model,
                    data_loader_val,
                    dataset_name=dataset_name,
                    iou_types=iou_types,
                    box_only=cfg.MODEL.RPN_ONLY,
                    device=cfg.MODEL.DEVICE,
                    expected_results=cfg.TEST.EXPECTED_RESULTS,
                    expected_results_sigma_tol=cfg.TEST.EXPECTED_RESULTS_SIGMA_TOL,
                    output_folder=output_folder,
                    skip_eval=args.skip_eval,
                    dllogger=dllogger,
                )
        synchronize()
        results.append(result)
    
    if is_main_process() and not args.skip_eval:
        map_results, raw_results = results[0]
        bbox_map = map_results.results["bbox"]['AP']
        segm_map = map_results.results["segm"]['AP']
        dllogger.log(step=tuple(), data={"BBOX_mAP": bbox_map, "MASK_mAP": segm_map})

if __name__ == "__main__":
    main()
    dllogger.log(step=tuple(), data={})