detection/train.py

data_root = '/root/autodl-tmp/ui_dataset'

import logging
import os
from collections import OrderedDict
import torch
from torch.nn.parallel import DistributedDataParallel
import random
import cv2

import detectron2.utils.comm as comm
from detectron2.checkpoint import DetectionCheckpointer, PeriodicCheckpointer
from detectron2.config import get_cfg
from detectron2.utils.visualizer import Visualizer
from detectron2.data import (
    datasets,
    MetadataCatalog,
    get_detection_dataset_dicts,
    build_detection_test_loader,
    build_detection_train_loader,
)
from detectron2.engine import default_argument_parser, default_setup, default_writers, launch
from detectron2.evaluation import (
    CityscapesInstanceEvaluator,
    CityscapesSemSegEvaluator,
    COCOEvaluator,
    COCOPanopticEvaluator,
    DatasetEvaluators,
    LVISEvaluator,
    PascalVOCDetectionEvaluator,
    SemSegEvaluator,
    inference_on_dataset,
    print_csv_format,
)
from detectron2.modeling import build_model
from detectron2.solver import build_lr_scheduler, build_optimizer
from detectron2.utils.events import EventStorage

from icecream import ic, install
install()
ic.configureOutput(includeContext=True, contextAbsPath=True)

logger = logging.getLogger("detectron2")


def visualize(dataset_name='valid_ui', num=4, iter=0):
    if not os.path.exists('./imgs'):
        os.mkdir('./imgs')
    metadata = MetadataCatalog.get(dataset_name)
    dataset = get_detection_dataset_dicts(dataset_name)

    for i, d in enumerate(random.sample(dataset, num)):
        img = cv2.imread(d["file_name"])
        visualizer = Visualizer(img[:, :, ::-1], metadata=metadata, scale=0.5)
        vis = visualizer.draw_dataset_dict(d)
        cv2.imwrite(f'./imgs/{iter}_{dataset_name}_{i}.png', vis.get_image()[:, :, ::-1])


def get_evaluator(cfg, dataset_name, output_folder=None):
    """
    Create evaluator(s) for a given dataset.
    This uses the special metadata "evaluator_type" associated with each builtin dataset.
    For your own dataset, you can simply create an evaluator manually in your
    script and do not have to worry about the hacky if-else logic here.
    """
    if output_folder is None:
        output_folder = os.path.join(cfg.OUTPUT_DIR, "inference")
    evaluator_list = []
    evaluator_type = MetadataCatalog.get(dataset_name).evaluator_type
    if evaluator_type in ["sem_seg", "coco_panoptic_seg"]:
        evaluator_list.append(
            SemSegEvaluator(
                dataset_name,
                distributed=True,
                output_dir=output_folder,
            )
        )
    if evaluator_type in ["coco", "coco_panoptic_seg"]:
        evaluator_list.append(COCOEvaluator(dataset_name, output_dir=output_folder))
    if evaluator_type == "coco_panoptic_seg":
        evaluator_list.append(COCOPanopticEvaluator(dataset_name, output_folder))
    if evaluator_type == "cityscapes_instance":
        return CityscapesInstanceEvaluator(dataset_name)
    if evaluator_type == "cityscapes_sem_seg":
        return CityscapesSemSegEvaluator(dataset_name)
    if evaluator_type == "pascal_voc":
        return PascalVOCDetectionEvaluator(dataset_name)
    if evaluator_type == "lvis":
        return LVISEvaluator(dataset_name, cfg, True, output_folder)
    if len(evaluator_list) == 0:
        raise NotImplementedError(
            "no Evaluator for the dataset {} with the type {}".format(dataset_name, evaluator_type)
        )
    if len(evaluator_list) == 1:
        return evaluator_list[0]
    return DatasetEvaluators(evaluator_list)


def do_test(cfg, model, storage=None):
    results = OrderedDict()
    for dataset_name in cfg.DATASETS.TEST:
        data_loader = build_detection_test_loader(cfg, dataset_name)
        evaluator = get_evaluator(
            cfg, dataset_name, os.path.join(cfg.OUTPUT_DIR, "inference", dataset_name)
        )
        results_i = inference_on_dataset(model, data_loader, evaluator)
        results[dataset_name] = results_i
        if comm.is_main_process():
            logger.info("Evaluation results for {} in csv format:".format(dataset_name))
            print_csv_format(results_i)
            # dump to storage, save to tensorboard
            if storage != None:
                for key, value in results_i.items(): # key = bbox / segm; value = {'AP': xx, 'APm': xx, ...}
                    logging.info(f'key value: {key}, {value}')
                    logging.info(f'key: {key}')
                    out_aps_dict = {}
                    for k, v in value.items():
                        k = dataset_name + '_' + k
                        out_aps_dict[k] = v
                        # print('**{k: v.item() for k, v in comm.reduce_dict(results_i).items()}\n', type(**{k: v.item() for k, v in comm.reduce_dict(results_i).items()}))
                    storage.put_scalars(**out_aps_dict)
    if len(results) == 1:
        results = list(results.values())[0]
    return results


def do_train(cfg, model, resume=False):
    model.train()
    optimizer = build_optimizer(cfg, model)
    scheduler = build_lr_scheduler(cfg, optimizer)

    checkpointer = DetectionCheckpointer(
        model, cfg.OUTPUT_DIR, optimizer=optimizer, scheduler=scheduler
    )
    start_iter = (
        checkpointer.resume_or_load(cfg.MODEL.WEIGHTS, resume=resume).get("iteration", -1) + 1
    )
    max_iter = cfg.SOLVER.MAX_ITER

    periodic_checkpointer = PeriodicCheckpointer(
        checkpointer, cfg.SOLVER.CHECKPOINT_PERIOD, max_iter=max_iter
    )

    writers = default_writers(cfg.OUTPUT_DIR, max_iter) if comm.is_main_process() else []

    # compared to "train_net.py", we do not support accurate timing and
    # precise BN here, because they are not trivial to implement in a small training loop
    data_loader = build_detection_train_loader(cfg)
    logger.info("Starting training from iteration {}".format(start_iter))
    with EventStorage(start_iter) as storage:
        for data, iteration in zip(data_loader, range(start_iter, max_iter)):
            storage.iter = iteration

            loss_dict = model(data)
            losses = sum(loss_dict.values())
            assert torch.isfinite(losses).all(), loss_dict

            loss_dict_reduced = {k: v.item() for k, v in comm.reduce_dict(loss_dict).items()}
            losses_reduced = sum(loss for loss in loss_dict_reduced.values())
            if comm.is_main_process():
                storage.put_scalars(total_loss=losses_reduced, **loss_dict_reduced)

            optimizer.zero_grad()
            losses.backward()
            optimizer.step()
            storage.put_scalar("lr", optimizer.param_groups[0]["lr"], smoothing_hint=False)
            scheduler.step()

            if (
                cfg.TEST.EVAL_PERIOD > 0
                and (iteration + 1) % cfg.TEST.EVAL_PERIOD == 0
                and iteration != max_iter - 1
            ):
                visualize('valid_ui', 5, iteration)
                visualize('train_ui', 5, iteration)
                do_test(cfg, model, storage)
                # Compared to "train_net.py", the test results are not dumped to EventStorage
                comm.synchronize()

            if iteration - start_iter > 5 and (
                (iteration + 1) % 20 == 0 or iteration == max_iter - 1
            ):
                for writer in writers:
                    writer.write()
            periodic_checkpointer.step(iteration)


def setup(args):
    """
    Create configs and perform basic setups.
    """
    cfg = get_cfg()
    cfg.merge_from_file(args.config_file)
    cfg.merge_from_list(args.opts)
    cfg.freeze()
    default_setup(
        cfg, args
    )  # if you don't like any of the default setup, write your own setup code
    return cfg


def main(args):
    cfg = setup(args)

    datasets.register_coco_instances("train_ui", {},
                                     f"{data_root}/train/_annotations.coco.json",
                                     f"{data_root}/train")
    datasets.register_coco_instances("train_dora_ui", {},
                                     f"{data_root.replace('ui_dataset', 'dora_dataset')}/train.json",
                                     f"{data_root.replace('ui_dataset', 'dora_dataset')}/train")
    datasets.register_coco_instances("test_ui", {},
                                     f"{data_root}/test/_annotations.coco.json",
                                     f"{data_root}/test")
    datasets.register_coco_instances("valid_ui", {},
                                     f"{data_root}/valid/_annotations.coco.json",
                                     f"{data_root}/valid")
    datasets.register_coco_instances("valid_dora_ui", {},
                                     f"{data_root.replace('ui_dataset', 'dora_dataset')}/val.json",
                                     f"{data_root.replace('ui_dataset', 'dora_dataset')}/train")
    print('done registering datasets')


    model = build_model(cfg)
    logger.info("Model:\n{}".format(model))
    if args.eval_only:
        DetectionCheckpointer(model, save_dir=cfg.OUTPUT_DIR).resume_or_load(
            cfg.MODEL.WEIGHTS, resume=args.resume
        )
        return do_test(cfg, model)

    distributed = comm.get_world_size() > 1
    if distributed:
        model = DistributedDataParallel(
            model, device_ids=[comm.get_local_rank()], broadcast_buffers=False
        )

    do_train(cfg, model, resume=args.resume)
    return do_test(cfg, model)


if __name__ == "__main__":
    args = default_argument_parser().parse_args()
    print("Command Line Args:", args)
    launch(
        main,
        args.num_gpus,
        num_machines=args.num_machines,
        machine_rank=args.machine_rank,
        dist_url=args.dist_url,
        args=(args,),
    )