其他开源notebook

$$其他开源notebook$$

高分开源notebook

案例一 mmcv

转载于

因为本比赛是离线大赛，所以所用依赖包需要在自己电脑上下载好，再进行上传

!pip install ../input/mmcvwhl/addict-2.2.1-py3-none-any.whl
!pip install ../input/mmdetection20-5-13/mmcv-0.5.1-cp37-cp37m-linux_x86_64.whl
!pip install ../input/mmdetection20-5-13/terminal-0.4.0-py3-none-any.whl
!pip install ../input/mmdetection20-5-13/terminaltables-3.1.0-py3-none-any.whl

!cp -r ../input/mmdetection20-5-13/mmdetection/mmdetection .

!mkdir -p mmdetection/data/Wheatdetection/annotations
!cp -r ../input/global-wheat-detection/test mmdetection/data/Wheatdetection/test
!cp -r ../input/global-wheat-detection/sample_submission.csv mmdetection/data/Wheatdetection/
!mkdir mmdetection/configs/wheatdetection

!cp ../input/mmdetfasterrcnn/config/config/faster_rcnn_r50_fpn_1x_coco_test.py mmdetection/configs/wheatdetection
!cp ../input/mmdetfasterrcnn/config/config/wheat_detection_test.py mmdetection/configs/_base_/datasets
!cp ../input/mmdetfasterrcnn/config/config/__init__.py mmdetection/mmdet/datasets
!cp ../input/mmdetfasterrcnn/config/config/wheat.py mmdetection/mmdet/datasets

cd mmdetection

!cp -r ../../input/mmdetection20-5-13/cocoapi/cocoapi .

cd cocoapi/PythonAPI

!make

!make install

!python setup.py install

import pycocotools

cd ../..

!pip install -v -e .

cd ../

import sys
sys.path.append('mmdetection') # To find local version

from mmdet.apis import init_detector, inference_detector, show_result_pyplot
import mmcv
from mmcv import Config
from mmdet.models import build_detector
from mmcv.runner import load_checkpoint
from mmcv.parallel import MMDataParallel
from mmdet.apis import single_gpu_test
from mmdet.datasets import build_dataloader, build_dataset

import pandas as pd
import os
import json

from PIL import Image

import torch

ef format_prediction_string(boxes, scores):
    pred_strings = []
    for j in zip(scores, boxes):
        pred_strings.append("{0:.4f} {1} {2} {3} {4}".format(j[0], j[1][0], j[1][1], j[1][2], j[1][3]))

    return " ".join(pred_strings)


def gen_test_annotation(test_data_path, annotation_path):
    test_anno_list = []
    for img in os.listdir(test_data_path):
        if img.endswith('jpg'):
            img_info = {}
            img_info['filename'] = img
            img_size = Image.open(os.path.join(test_data_path, img)).size
            img_info['width'] = img_size[0]
            img_info['height'] = img_size[1]
            test_anno_list.append(img_info)
    with open(annotation_path, 'w+') as f:
        json.dump(test_anno_list, f)

DIR_INPUT = '/kaggle/working/mmdetection/data/Wheatdetection'
DIR_TEST = f'{DIR_INPUT}/test'
DIR_ANNO = f'{DIR_INPUT}/annotations'

DIR_WEIGHTS = '/kaggle/input/mmdetfasterrcnn'
WEIGHTS_FILE = f'{DIR_WEIGHTS}/epoch_50.pth'

test_df = pd.read_csv(f'{DIR_INPUT}/sample_submission.csv')

# prepare test data annotations
gen_test_annotation(DIR_TEST, DIR_ANNO + '/detection_test.json')

config_file = '/kaggle/working/mmdetection/configs/wheatdetection/faster_rcnn_r50_fpn_1x_coco_test.py'
cfg = Config.fromfile(config_file)
cfg.data.test.test_mode = True

distributed = False

dataset = build_dataset(cfg.data.test)
data_loader = build_dataloader(
    dataset,
    samples_per_gpu=1,
    workers_per_gpu=1,
    dist=distributed,
    shuffle=False)

#################################### faster rcnn ############################################
device = torch.device('cuda') if torch.cuda.is_available() else torch.device('cpu')

model = build_detector(cfg.model, train_cfg=None, test_cfg=cfg.test_cfg)
checkpoint = load_checkpoint(model, WEIGHTS_FILE, map_location='cpu')

model.CLASSES = dataset.CLASSES

model = MMDataParallel(model, device_ids=[0])
outputs = single_gpu_test(model, data_loader, False, None, 0.5)

results = []

for images_info, result in zip(dataset.data_infos, outputs):
    boxes = result[0][:, :4]
    scores = result[0][:, 4]

    boxes[:, 2] = boxes[:, 2] - boxes[:, 0]
    boxes[:, 3] = boxes[:, 3] - boxes[:, 1]

    result = {
        'image_id': images_info['filename'][:-4],
        'PredictionString': format_prediction_string(boxes, scores)
    }

    results.append(result)

test_df = pd.DataFrame(results, columns=['image_id', 'PredictionString'])

# save result
test_df.to_csv('submission.csv', index=False)

!rm -rf mmdetection/