Collect and Label Images to Train a YOLOv5 Object Detection Model in PyTorch | Part 2
Hey everyone and welcome to the second part of the YOLOv5 series!
In this video (Part 2) we'll learn about collecting and labelling data in order to train a YOLOv5 model on a custom dataset.
We'll start by asking if object detection is the right approach to begin with. If it is, we'll talk about creating your own dataset, using labelling tools to draw bounding boxes, and using Weights & Biases Artifacts to upload data to the cloud, so that we can train on a virtual machine like Google Colab.
Links:
📍 Example Colab for training: https://wandb.me/yolo-colab
📍 ModifiedOpenLabelling: https://github.com/ivangrov/ModifiedO...
📍 OpenLabelling: https://github.com/Cartucho/OpenLabeling
📍 COCO dataset: https://cocodataset.org/#home
📍Video I made about downloading Google Images: • Web Scraping (Downloading) Google Ima...
Upload dataset command (different than what I show in the video, slightly updated): python utils/loggers/wandb/log_dataset.py --project custom_yolov5 --data data/custom_dataset.yaml
---
You can watch the whole YOLOv5 Series here:
🚀Part 0 - Overview of the YOLOv5 and W&B integration: • Train and Debug YOLOv5 Models with We... Rocket
🚀Part 1 - Install YOLOv5 on Windows and Google Colab: • Install YOLOv5 to Detect Objects on W... Rocket
🚀Part 2 - Collect & Label a Custom Dataset: • Collect and Label Images to Train a Y... Rocket
🚀Part 3 - Train a Custom YOLOv5 Model to Detect Bus Numbers: • Train a Custom YOLOv5 Model on Google...
---
Follow Ivan:
👉 Twitter: / ivangrov
👉 YouTube: / ivangoncharovai
---
⏳ Timestamps ⏳
0:00 Intro & overview
1:00 Is object detection right for your problem?
2:58 Creating a dataset
5:01 The importance of true negatives
6:13 The YOLOv5 label format
8:03 Labelling tools and labelling
12:05 Using W&B Artifacts to upload a dataset to the cloud
14:51 Using W&B Tables to visualize a dataset in the browser
16:51 Outro
---
Join our community of ML practitioners where we host AMAs, share interesting projects and meet other people working in Deep Learning: http://wandb.me/slack
Check out Fully Connected, which features curated machine learning reports by researchers exploring deep learning techniques, Kagglers showcasing winning models, industry leaders sharing best practices, and more: https://wandb.ai/fully-connected
--
Credits
📍 VBTI (stuffed animal inside a bounding box): / carlolepelaars_vbti-ai-deeplearning-activi...
📍 3D Object Detection: • Monocular Quasi Dense 3D Object Track...
📍 Image segmentation: • ICNet for Real-Time Semantic Segmenta...