F1 CAR DETECTION
[Baseline] F1 Detection
Baseline notebook for F1 Detection Challenge of Blitz 8
Shubhamaicrowd
Getting Started Code for F1 Car Detection Challenge on AIcrowd¶
Author : Shubhamai¶
Download Necessary Packages 📚¶
# Installing the AIcrowd CLI
!pip install aicrowd-cli
# Installing PyTorch
!pip install pyyaml==5.1
!pip install torch==1.7.1 torchvision==0.8.2
import torch, torchvision
print(torch.__version__, torch.cuda.is_available())
!gcc --version
# Installing Detectron2
import torch
assert torch.__version__.startswith("1.7")
!pip install detectron2 -f https://dl.fbaipublicfiles.com/detectron2/wheels/cu101/torch1.7/index.html
Download Data ⏬¶
The first step is to download out train test data. We will be training a model on the train data and make predictions on test data. We submit our predictions.
API_KEY = "6c4dc1818e928ae4f5ca02323114f662" #Please enter your API Key from [https://www.aicrowd.com/participants/me]
!aicrowd login --api-key $API_KEY
!aicrowd dataset download --challenge f1-car-detection -j 3
Below, we create a new directory to put our downloaded data! 🏎
We unzip the ZIP files and move the CSVs.
!rm -rf data
!mkdir data
!unzip train.zip -d data/train > /dev/null
!unzip val.zip -d data/val > /dev/null
!unzip test.zip -d data/test > /dev/null
!mv train.csv data/train.csv
!mv val.csv data/val.csv
Import packages 📦¶
It's time to import all the packages that we have downloaded and the packages that we will be needing for building our model.
import torch, torchvision
import detectron2
from detectron2.utils.logger import setup_logger
setup_logger()
from detectron2 import model_zoo
from detectron2.engine import DefaultTrainer, DefaultPredictor
from detectron2.config import get_cfg
from detectron2.utils.visualizer import Visualizer
from detectron2.data import MetadataCatalog, DatasetCatalog
from detectron2.structures import BoxMode
import pandas as pd
import numpy as np
import os
from PIL import Image, ImageDraw
import matplotlib.pyplot as plt
from tqdm.notebook import tqdm
import cv2
import random
from ast import literal_eval
data_path = "data"
train_df = pd.read_csv(os.path.join(data_path, "train.csv"))
val_df = pd.read_csv(os.path.join(data_path, "val.csv"))
Visualize the data 👀¶
Using Pandas and the Matplot Library in Python, we will be viewing the images in our datasets.
train_df
# Defining a function to take a look at the images
def show_images(images, num = 5):
images_to_show = np.random.choice(images, num)
for image_id in images_to_show:
image = Image.open(os.path.join(data_path, f"train/{image_id}.jpg"))
bbox = literal_eval(train_df.loc[train_df['ImageID'] == image_id]['bboxes'].values[0])
draw = ImageDraw.Draw(image)
draw.rectangle([bbox[0], bbox[2], bbox[1], bbox[3]], width=1)
plt.figure(figsize = (15,15))
plt.imshow(image)
plt.show()
show_images(train_df['ImageID'].unique(), num = 5)
Creating Dataset 🎈¶
In the section below, we will be creating the dataset that will be put into our Model for training!
dict_dataset = []
def get_dataset_dics():
for index, row in train_df.iterrows():
image = Image.open(os.path.join(data_path, f"train/{row['ImageID']}.jpg"))
w, h = image.size
ann_lst = []
bbox = literal_eval(row['bboxes'])
ann_dict = {'bbox': [bbox[0], bbox[2], bbox[1], bbox[3]],
'bbox_mode': BoxMode.XYXY_ABS,
'category_id': 0, #i[1]['category_id'].values[0],
'iscrowd': 0}
ann_lst.append(ann_dict)
image_dict = {'annotations': ann_lst,
'file_name': os.path.join(data_path, f"train/{row['ImageID']}.jpg"),
'height': h,
'image_id': row["ImageID"], #i[1]['image_category_id'].values[0],
'width': w}
dict_dataset.append(image_dict)
return dict_dataset
dict_dataset = get_dataset_dics()
d = f"f1_train{np.random.randint(10000)}"
DatasetCatalog.register(d, lambda d=d : get_dataset_dics())
MetadataCatalog.get(d).set(thing_classes=["F1Cars"])
obj_metadata = MetadataCatalog.get(d)
for i in random.sample(dict_dataset, 3):
img = cv2.imread(i["file_name"])
visualizer = Visualizer(img, metadata=obj_metadata, scale=0.5)
out = visualizer.draw_dataset_dict(i)
plt.imshow(out.get_image())
Creating the Model 🏎¶
Now that we have the dataset is ready, it's time to create a model that we will train on our data!
cfg = get_cfg()
cfg.merge_from_file(model_zoo.get_config_file("COCO-Detection/faster_rcnn_R_50_DC5_3x.yaml"))
cfg.DATASETS.TRAIN = (d,)
cfg.DATASETS.TEST = ()
cfg.DATALOADER.NUM_WORKERS = 2
cfg.MODEL.WEIGHTS = model_zoo.get_checkpoint_url("COCO-Detection/faster_rcnn_R_50_DC5_3x.yaml")
cfg.SOLVER.IMS_PER_BATCH = 2
cfg.SOLVER.BASE_LR = 0.00025
cfg.SOLVER.MAX_ITER = 500
cfg.MODEL.ROI_HEADS.BATCH_SIZE_PER_IMAGE = 128
cfg.MODEL.ROI_HEADS.NUM_CLASSES = 1
os.makedirs(cfg.OUTPUT_DIR, exist_ok=True)
Train the Model 🏃🏽♂️¶
trainer = DefaultTrainer(cfg)
trainer.resume_or_load(resume=False)
trainer.train()
Let's take a look at the loss and several other metrics by running the pice of code below:
%load_ext tensorboard
%tensorboard --logdir output
Testing Phase 😅¶
We are almost done. We trained and validated on the training data. Now its the time to predict on test set and make a submission.
Loading Pretrained Model¶
Before we predict on any data, let's quickly reload the model to predict on our dataset.
cfg.MODEL.WEIGHTS = os.path.join(cfg.OUTPUT_DIR, "model_final.pth")
cfg.MODEL.ROI_HEADS.SCORE_THRESH_TEST = 0.5
predictor = DefaultPredictor(cfg)
Predict Test Set¶
Predict on the test set and you are all set to make the submission!
test_imgs_paths = os.listdir(os.path.join(data_path, "test"))
predictions = {"ImageID":[], "bboxes":[]}
for test_img_path in tqdm(test_imgs_paths):
img = cv2.imread(os.path.join(data_path, "test", test_img_path))
h, w, _ = img.shape
model_predictions = predictor(img)
bboxes = model_predictions['instances'].pred_boxes.tensor.cpu().numpy().tolist()
scores = model_predictions['instances'].scores.cpu().numpy().tolist()
for n, bbox in enumerate(bboxes):
bbox.append(scores[n])
bboxes[n] = bbox
image_id = test_img_path.split('.')[0]
predictions['ImageID'].append(image_id)
predictions['bboxes'].append(bboxes)
Save the prediction to csv¶
We have the DataFrame for our predictions are ready for submission. Now, we will make a submission to the AIcrowd platform.
You can directly upload the CSV to the challenge or use the AIcrowd CLI to make a final submission directly from this Colab Notebook.
submission = pd.DataFrame(predictions)
submission
submission.to_csv("submission.csv", index=False)
Making Direct Submission¶
Submitting with a single line of code.
!aicrowd submission create -c f1-car-detection -f submission.csv
Content
Comments
You must login before you can post a comment.
how do i transfer this data to colab , as i don’t have enough internet to download the data.. can you help me with the instruction please