LightToYang

Is it possible training SSD or YOLOv3 in fp16?

13

I find that in train_imagenet.py --dtype could be set float16 for faster training, `parser.add_argument('--dtype', type=str, default='float32', help='data type for training. default is float32')` but I don't the similar setting in...

The difference of RFBNet in RFB_Net_vgg.py and Fig.5 ?

9

Hello, thank you for code releasing. There are two branches from fc7 layer output in Fig.5, one is the input of RBF and another is the input of RBF(stride 2)....

Can this work extend to other domains like segmentation/detection ?

1

It seems there is no special limitation for this work.

Image format converting questions

6

I am working on speeding up image loading for training models on ImageNet Dataset. Based on https://stackoverflow.com/questions/57663734/how-to-speed-up-image-loading-in-pillow-python, ``` #!/usr/bin/env python3 import numpy as np import pyvips import cv2 from PIL...

It seems the annotations are incomplete for some pictures?

1

  

nvidia-device-plugin is starting but still no gpu-device in node description and gpu-pod FailedScheduling

4

### 1. Issue or feature description nvidia-device-plugin is starting but still no gpu-device in node description and gpu-pod FailedScheduling ### 2. Steps to reproduce the issue After `kubectl create -f...

how to replace 3ddfa_v2 with 3ddfa ?

2

3ddfa_v2 is much accuracy than 3ddfa on face fitting

根据TSM部署教程获得错误的结果

1

https://github.com/PaddlePaddle/PaddleVideo/tree/develop/deploy/python_serving 期望结果 `# rpc方式打印的结果 PipelineClient::predict pack_data time:1645631086.764019 PipelineClient::predict before time:1645631086.8485317 key: "label" key: "prob" value: "[\'archery\']" value: "[0.9907388687133789]" ` 然而 `PipelineClient::predict pack_data time:1681895941.1273232 PipelineClient::predict before time:1681895941.1585848 key: "label" key: "prob"...

the predictions.json by `model.predict(img_path, save=False, conf=0.001)` get worse LVIS-minival results ``` Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=300 catIds=all] = 0.206 Average Precision (AP) @[ IoU=0.50 |...