β What is Multi-Scale Training (MST)?
π‘ MTS helps your model to be robust to image sizes, an get better performance
β’ Training on small images is faster
β’ Training on large images increases your model performance
How is MST done?
π‘ MTS helps your model to be robust to image sizes, an get better performance
β’ Training on small images is faster
β’ Training on large images increases your model performance
How is MST done?
Every N (e.g., 10) epochs, we randomly chooses a new image dimension from a range of sizes [640, 768, 800], and train our model
This means the same network becomes better at predicting at different resolutions.
This means the same network becomes better at predicting at different resolutions.
β’ In MMDetetection, models trained using multi-scale technique have β_mstrain_β in their name.
β’ Example: vfnet_r50_fpn_mstrain_2x_coco
β’ Example: vfnet_r50_fpn_mstrain_2x_coco
π‘ MTS is a bit different than Progressive Resizing Training (PRT)
β’ In PRT, we first train our model using a small image size (e.g. 640) for X epochs,
β’ Then, we gradually increase the size: 640 -> 768 -> 800
β’ In PRT, we first train our model using a small image size (e.g. 640) for X epochs,
β’ Then, we gradually increase the size: 640 -> 768 -> 800
β If you are interested in mastering Object Detection (OD), follow @ai_fast_track, to receive highly curated content right in your feed.
β I started a newsletter, join to master object detection and get a competitive edge in this field
π° Newsletter: getrevue.co/profile/ai_fasβ¦
β I started a newsletter, join to master object detection and get a competitive edge in this field
π° Newsletter: getrevue.co/profile/ai_fasβ¦
β’ β’ β’
Missing some Tweet in this thread? You can try to
force a refresh
γ
