Hongyang Li
Hongyang Li
Hi, the core lies in the pre-trained model, which is trained on the meta training set. This is common practice as in [LEO](https://arxiv.org/pdf/1807.05960.pdf), [Qiao et al](https://arxiv.org/abs/1706.03466), for example. Unfortunately, the...
Hi, a quick reply: please refer to the paper config. The 20way_1shot.yaml is a bad example. I would fix it asap. On Thu, Sep 26, 2019 at 1:31 AM Bharti...
Hi! Thanks for the interest! Yeah, in preliminary code there is a mask branch indeed. I will update this semi-pulic-ready codebase very soon.
Hmm..that's weird. How do you split the reliable (large) /less reliable (small) sets? Similarly as we do for the anchors? The losses (cls, bbox, meta; ignore the mask loss) are...
> @RicardLee Hello, thanks for your response. As of now, I simply wanted to use your pipeline to create 3d lanes on existing datasets. But I have no doubts that...
Caffe with Matlab interface
got the same issue
problem solved. because some conv and BN layers are not initialized. (`train-cifar.lua` and `residual-layers.lua`). I spent a whole day debugging why the loss does not decrease. (the net just randomly...
Nope. As we plan to host the Challenge in 2024 with some modifications, releasing the procedure of generating annotations is definitely bummer no. You are welcome to contribute! Thanks