-
According to figure 3 of the BiFPN Layer in the paper, we take the last feature for each resolution as output.
Does it lose a break statement after line 533 in efficientdet_arch.py?
https://github.c…
-
Hello, I got a question about GPU memory usage. I train a model with 46 class and head only false on d7. GPU is Tesla P100 16GB. I got out of memory even for batchsize=1. What GPU do you use for train…
-
I want to finetune only the last layers of EfficientNet and BiFPN, instead of the current whole net or head only, could you give some advice?
Thank you for your excellent work.
-
**CSPNet**: A New Backbone that can Enhance Learning Capability of CNN
The most comprehensive comparison of detection models:
* paper: https://arxiv.org/abs/1911.11929v1
* models: https://github.c…
-
I haven’t found about it is two different tensors in tensorflow version. Thanks
-
This is a question, not a bug. @AlexeyAB is there a way to incorporate the "Visually Coherent Image Mixup" augmentation strategy? Amazon recently published a paper that claims that this approach resul…
-
Hey! Thanks for your awesome work!
I've been trying to train a custom b0 model on my own data. I used the following two commands for step 1 and 2:
python3 train.py --snapshot imagenet --phi 0 --…
-
Hello, everyone
I tried to train the efficientDet-d5 model with config
```
label_smoothing: 0.1
learning_rate: 0.008
lr_warmup_init: 0.0008
max_instances_per_image: 1000
mixed_precision: …
-
`def forward(self, inputs):
max_size = inputs.shape[-1]
_, p3, p4, p5 = self.backbone_net(inputs)
features = (p3, p4, p5)
features = self.bifpn(features)
…
-
Thanks for your implementation, I wanted to modify your code to make it as segmentation network as suggested in the paper. We have to
> Following [16], we modify our EfficientDet model to keep fea…