As you may notice, the model does not have a scales params in Resize....

Does anyone knows why it does needs scales but onnx opset 10 said, Resize op consume 2 inputs which is input nodes and scales.

Compares to yolov3.onnx, much more standared:

![image](https://user-images.

if cfg.TRAIN.RPN_POSITIVE_WEIGHT < 0:
            num_examples = torch.sum(labels[i] >= 0)
            positive_weights = 1.0 / num_examples
            negative_weights = 1.0 / num_examples
        else:
            assert ((cfg.TRAIN.RPN_POSITIVE_WEIGHT > 0) &
                    (cfg.TRAIN.RPN_POSITIVE_WEIGHT < 1))

In above code

1. in line 2, there is no definition for index i

Are there any changes that I need to make in order to train a dataset with such images?
Side note: the images are also 4084x3xxx, I would also like to know what sort of changes I need to make to accomodate my images.

The 2x down-sampling is one of the important operations in reference models. But, a convolution or a pooling with stride=2, padding='SAME' may result in different outputs over different deep learning libraries (e.g., TensorFlow, CNTK, Theano, Caffe, Torch, ...) due to their different padding behaviors.

For example (TensorNets syntax; but can be regarded as pseudo codes for other libraries),

I have followed the steps to run demo but I get inaccurate results. The boxes of heads are seen to be offset from their original coordinate(see the results).

In convert_tfrecords.py, it converts all images in VOCdevkit\VOC2012\JPEGImages into tfrecords for training,
https://github.com/HiKapok/SSD.TensorFlow/blob/b47ff6164c8925a8bbccc593719d5bbbab996058/dataset/convert_tfrecords.py#L366-L370

However, not all the images in VOC2012 are for detection, there are only 11540 images specified in ImageSets\Main\trainval.txt for detection. But in JPEGImages

Issue summary

Cannot run demo

Steps to reproduce

When I try to run the command sh example/FRCNN/demo_frcnn.sh it says:

is empty, default is 0
examples/FRCNN/demo_frcnn.sh: line 14: build/examples/FRCNN/demo_frcnn_api.bin:No such file or directory

Your system configuration

Operating system: windows 10
Compiler: MinGW64
CUDA version (if applicable): 9.0
CUDN

这个im_info中的scale应该是指图像卷积完后的尺寸/输入时图像resize的固定尺寸。
但是我看了paddledetection源码和aistu上的案例，im_info都是输入时图像resize的固定尺寸/图像原来的尺寸，举一个例子
图像原来的尺寸：1320
输入时图像resize的固定尺寸：1300
图像卷积完后的尺寸：1300/16=86.25
![image](https://user-images.githubusercontent.com/43198176/73636004-b47c2280-469f-11ea-836c-309c504f74