Hi, I'd like to ask that if the performance on cityscapes reported in the model_zoo/README.md is the trained using both fine and coarse labeled data? I tried to run your code using only fine labeled data at s=1.5, 100 epochs for stage1 and 100 epochs for stage2 but the best mIoU on val is 61.1%, 2.7% lower than your implementation. I wonder if it's because I didn't train enough or the data problem.

Thank you for sharing this great code. Right row, I want to deploy your model in to tvm platform, which may need conversion between pytorch and onnx, the code I used is like below.

weights = 'model/detection/model_zoo/espnetv2/espnetv2_s_2.0_pascal_300x300.pth' model = ssd(args, cfg) pretrained_dict = torch.load(weights, map_location=torch.device('cpu')) model.load_state_dict(pretrained_dict) PATH_ONNX='deploy.onnx' dummy_input = torch.randn(1, 3, 300, 300, device='cpu') torch.onnx.export(model, dummy_input, PATH_ONNX, input_names = ['image'], output_names= ['output'], verbose=True,opset_version=11)

but during the conversion, an error occurs,the info is below:

~/software/EdgeNets/nn_layers/eesp.py:139: TracerWarning: Converting a tensor to a Python boolean might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs! if w2 == w1: ~/software/EdgeNets/nn_layers/eesp.py:89: TracerWarning: Converting a tensor to a Python boolean might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs! if expanded.size() == input.size(): ~/software/EdgeNets/nn_layers/efficient_pyramid_pool.py:44: TracerWarning: Converting a tensor to a Python float might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs! h_s = int(math.ceil(height * self.scales[i])) ~/software/EdgeNets/nn_layers/efficient_pyramid_pool.py:45: TracerWarning: Converting a tensor to a Python float might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs! w_s = int(math.ceil(width * self.scales[i]))

raise RuntimeError("Failed to export an ONNX attribute, " RuntimeError: Failed to export an ONNX attribute, since it's not constant, please try to make things (e.g., kernel size) static if possible

please give some tips, which I can figure out the problem. thank you for your help!

In the paper 《DiCENet: Dimension-wise Convolutions for Efficient Networks》, the network width scaling parameter s can be selected, but in the experiment of image multi-label classification, the experimental error of s=0.1 (when I run the corresponding program of s=0.2, the machine can't run, but the network parameter of s=0.1 is less than half) ... can you provide a program for s = 0.1?

Hello guys, after reading your article on ESPNETv2 I was very excited about your optimizations in convolutional operations, and would very much like to make the network run on a raspberry pi.

After much effort, I was able to compile all the necessary dependencies on raspberry pi, and was able to successfully execute the detection_demo.py script, which by default makes inferences on the images in your repository. However, when I use another set of images (passed as a parameter in --im-dir) the network ends up running with the following message: Segmentation Fault.

Do you have any idea what might be generating this? I tried to use several other image sets, and the only one the network could run without any problems was the default image set from its repository.

I am using the raspberry pi 3 model B + with the Raspibian Buster Lite operating system.

1. For multi classification, while calculating loss for coco dataset why it is multiplied by number of classes (80.0)? Is it weight parameter for class imbalance? loss = criteria(output, target.float()) * 80.0

2. For calculating precision and recall should we use cumulative TP & FP as mentioned here: https://github.com/rafaelpadilla/Object-Detection-Metrics

I have tried to convert pytorch model to onnx with api (torch.onnx.export), but I meet a problem showing below info Failed to export an ONNX attribute, since it's not constant, please try to make things (e.g., kernel size) static if possible How could I fix it ?

Hi, thanks for publishing the code.

I noticed the optimizer.zero_grad() is called after calling optimizer.step() in train_eval_seg.py. Is this intentional? I was having issues until I moved zero_grad() above loss.backward().

Also when resuming the training there was a runtime error with type compatibility (cpu vs cuda) when calculating loss.backward() and adding model = model.cuda() right after loading the state dictionary in train_segmentation.py solved it.

The training script would stop throwing an error about multiplying Bool with ByteTensor in segmentation_miou.py at

        pred = pred * (target>0)
        inter = pred * (pred == target)

I assume the authors were expecting torch to cast Bool values into 0 and 1, which didn't happen. I changed it to

        pred = pred * (target>0).type(torch.ByteTensor)
        inter = pred * .type(torch.ByteTensor)

to explicitly cast the values.

I'm not sure why this happened in my case, perhaps due to version incompatibility.

Hi I got the follow error when i convert espnetv2 segment model to onnx. raise RuntimeError("Failed to export an ONNX attribute, " RuntimeError: Failed to export an ONNX attribute, since it's not constant, please try to make things (e.g., kernel size) static if possible

how i can fix it?

thanks for the great work i would like to use the idea of DiceNet on a custom dataset. And i'm new to DL and pytorch can you give me some idea how to make it?

Thank you:)

Sorry for disturbing you. When I run the test_detection.py ,I got an error:RuntimeError: expected backend CUDA and dtype Float but got backend CPU and dtype Float. How can I solve it? Thank you

Hi, I tried EgeNets in object detections. It is working fine with objects. My question is about the walls and surfaces (floors and ceilings). Are there more properties to be set to make the model detect them I mean walls and surfaces.

thanks, Dina

Update to ensure that different numbers of input channels can be handled without error. (Shortcut connection configuration assumed input channel count always equal to config_inp_reinf.)

The following are the correct lines in train_detection.py to enable resume from checkpoint.

        checkpoint = torch.load(args.resume, map_location=torch.device('cpu'))
parser.add_argument('--resume', type=str, help='resume from checkpoint') #action='store_true',

@sacmehta Really impressive work!

It is great to see that you have open-sourced the implementation of DimConv😁

We are wondering whether you have any plan to release the optimized implementation illustrated in the following Figures:

In the input reinforcement branches it seem that there are redundants Average pool on the original input in the deeper stages. Do you see any problem to pass the strided average input image after each DownSample block to the next one? In DownSampler forward method, return input2:

def forward(self, input, input2=None):
        '''
        :param input: input feature map
        :return: feature map down-sampled by a factor of 2
        '''
        avg_out = self.avg(input)
        eesp_out = self.eesp(input)
        output = torch.cat([avg_out, eesp_out], 1)

        if input2 is not None:
            #assuming the input is a square image
            # Shortcut connection with the input image
            w1 = avg_out.size(2)
            while True:
                input2 = F.avg_pool2d(input2, kernel_size=3, padding=1, stride=2)
                w2 = input2.size(2)
                if w2 == w1:
                    break
            output = output + self.inp_reinf(input2)

        return self.act(output), input2

In Espnet class, overwrite the input object:

out_l3_0, input = self.level3_0(out_l2, input)  # down-sample