Skip to main content
W&B Fully Connected
>
Articles
>
Object Detection

Object Detection with PaddleDetection and Weights & Biases

In this article, we'll learn how to use PaddleDetection to train a YOLOX model from scratch, logging all of your metrics and model checkpoints with Weights & Biases.
Manan Goel
Created on May 24|Last edited on February 2

runs.summary["Predictions"]
 - 3 of 7
Input Image
Annotated Image
2
3
Run set
16

PaddleDetection is an end-to-end object detection development kit based on PaddlePaddle. It implements varied mainstream object detection, instance segmentation, tracking, and keypoint detection algorithms in modular design with configurable modules such as network components, data augmentations, and losses.
PaddleDetection now comes with a built-in W&B integration that logs all your training and validation metrics, your model checkpoints, and their corresponding metadata.
If you would like to follow along this tutorial with running code, check out the colab link below!
💡
In this tutorial, we'll train a YOLOX model on a subset of the COCO2017 dataset, which contains 1000 images in the training and 250 in the validation set.
You can activate the W&B logger for your training jobs in two ways:
  • Command Line: The arguments to the W&B logger must be proceeded by -o, and each individual argument must contain the prefix "wandb-" and the `--use_wandb` flag should be used.
python tools/train -c config.yml --use_wandb -o wandb-project=MyDetector wandb-entity=MyTeam wandb-save_dir=./logs
  • YAML File: Add the arguments to the configuration YAML file under the wandb header like this
wandb:
  project: MyProject
  entity: MyTeam
  save_dir: ./logs

Setting Things Up

Installing the W&B SDK

We start by installing the wandb SDK, followed by logging into our W&B account:
pip install wandb
wandb login

Installing PaddleDetection

Next, we'll clone the PaddleDetection library and install the package from the source since it also contains a training script along with configurations for a wide variety of pre-implemented models!
pip install paddlepaddle-gpu pyclipper attrdict gdown -qqq
git clone https://github.com/PaddlePaddle/PaddleDetection
cd PaddleDetection
pip install -e .

How to Download the COCO Dataset

We begin by downloading a subset of the COCO 2017 dataset created using this amazing repository.
The dataset has been logged as a W&B Artifact for easier downloading. It contains 1000 images for training and 250 for validation with corresponding annotations, which we will now use for training our object detection model.
artifact = wandb.Api().artifact("manan-goel/PaddleDetectionYOLOX/COCOSubset:latest")
path = artifact.download(root='./dataset/coco')

Training

Training the Model

We now use the training script in the PaddleDetection library to train the YOLOX model, and the config above adds W&B logging during training. We also add the --eval flag to have an evaluation step every 5 epochs.

Using the CLI

python tools/train.py -c configs/yolox/yolox_nano_300e_coco.yml --use_wandb -o wandb-project=PaddleDetectionYOLOX --eval

Using the Configuration YAML File

In order to automatically start experiment tracking with W&B for your training pipelines, the following snippet can be added to the configuration YAML file, which is then used as input to the training script.
wandb:
  project: PaddleDetectionYOLOX
We will edit this file for our usage in this section.

Visualization

Tracking the Training and Validation Metrics

During training, the metrics on the training and validation sets are logged to a W&B dashboard which looks something like this:

Run set
16


Tracking System Metrics


Run set
16


Downloading the Best Model From W&B

The model checkpoints are logged as W&B artifacts and can be downloaded for evaluation using the following snippet.
import wandb
artifact = wandb.Api().artifact('manan-goel/PaddleDetectionYOLOX/model-26oqc38r:best', type='model')
artifact_dir = artifact.download()

Testing on Images

The following cell runs the inference script in the PaddleDetection repository on all the images in the demo directory. It stores them in the infer_output directory using the YOLOX model pulled from W&B.
for i in $(ls demo/*.jpg)
do
    python tools/infer.py -c configs/yolox/yolox_nano_300e_coco.yml \
                    --infer_img=$i \
                    --output_dir=infer_output/ \
                    --draw_threshold=0.5 \
                    -o weights=./artifacts/model-26oqc38r:v1/model
done
This goes through each jpg file in the demo directory and runs the inference script on them using the downloaded model, and stores the annotated images in the infer_output directory.

Bonus: Logging Annotated Images to Your W&B Dashboard

import glob
import wandb

wandb.init(project="PaddleDetectionYOLOX")
wandb.use_artifact('manan-goel/PaddleDetectionYOLOX/model-26oqc38r:best')
table = wandb.Table(columns=["Input Image", "Annotated Image"])

inp_imgs = sorted(glob.glob("./demo/*.jpg"), key=lambda x: x.split("/")[-1])
out_imgs = sorted(glob.glob("./infer_output/*.jpg"), key=lambda x: x.split("/")[-1])

for inp in inp_imgs:
    for out in out_imgs:
        if out.split("/")[-1] != inp.split("/")[-1]:
            continue
        table.add_data(
            wandb.Image(inp),
            wandb.Image(out)
        )

wandb.log({
    "Predictions": table
})
wandb.finish()
This above script will take the input images and the labeled images and log them to a W&B table which looks something like this:

Run set
16


Conclusion

This tutorial gives a quick run-through on how you can use W&B in conjunction with PaddleDetection to support all your object detection model development needs. Check out the Colab for a version of this report with executable code.

Tags: Object Detection, COCO, YOLO, Computer Vision, Tutorial, Intermediate, Articles

Created with ❤️ on Weights & Biases.

https://wandb.ai/manan-goel/PaddleDetectionYOLOX/reports/Object-Detection-with-PaddleDetection-and-Weights-Biases--VmlldzoyMDU4MjY0
Made with Weights & Biases. Sign up or log in to create reports like this one.
Iterate on AI agents and models faster. Try Weights & Biases today.
Sign upTry W&B now
List<Maybe<File<(table)>>>