Train with random weights

[sh7jacobi]

How I can train without pretrained weights?

[fyangneil]

pretrained model? You mean the model trained on ImageNet？ You can download it in https://github.com/facebookresearch/Detectron/blob/master/MODEL_ZOO.md [https://avatars.githubusercontent.com/u/16943930?s=400&v=4]https://github.com/facebookresearch/Detectron/blob/master/MODEL_ZOO.md Detectron/MODEL_ZOO.md at master · facebookresearch/Detectron · GitHubhttps://github.com/facebookresearch/Detectron/blob/master/MODEL_ZOO.md Notes: A deeper backbone architecture is used: ResNeXt-152-32x8d-FPNThe backbone ResNeXt-152-32x8d model was trained on ImageNet-5k (not the usual ImageNet-1k)Training uses multi-scale jitter over scales {640, 672, 704, 736, 768, 800} github.com

---

From: sh7jacobi [email protected] Sent: Friday, January 29, 2021 10:33 AM To: fyangneil/Clustered-Object-Detection-in-Aerial-Image [email protected] Cc: Subscribed [email protected] Subject: [External] [fyangneil/Clustered-Object-Detection-in-Aerial-Image] Train with random weights (#16)

How I can train without pretrained weights?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHubhttps://github.com/fyangneil/Clustered-Object-Detection-in-Aerial-Image/issues/16, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ALAUWTUADGEIK2MNGDCJJGLS4LIM3ANCNFSM4WZBND6A.