Sam_S

60 might be too small of a sample. I trained with around ~1200 images and it was better. If you have fewer images, train with fewer epochs and use a...

This was a lifesaver honestly. Nothing else was working for me except explicitly asking for the android permissions.

In https://github.com/cheind/py-motmetrics/issues/48, when only observations with label ID=1 (Pedestrian) in the MOT16-02 challenge is used, inputting ground truth as the test set gives expected results. Also, @cheind, I checked and...

Add this in the training code ```python from torch.utils.tensorboard import SummaryWriter tboard_writer = SummaryWriter(log_dir="experiments") # to add the network structure in the tensorboard logger # make sure the type matches...

`c, h, w` refer to the input channel, height and width of network. Based on what network you are using these values might be `3, 320, 320`. Check the `config/yaml`...

This line `tboard_writer = SummaryWriter(log_dir="experiments")` should create an experiments folder in your working directory where tensorboard stores its log files. Then from a different terminal, you can enter `tensorboard --logdir=experiments...

The author has disabled the color parameter in the `draw_boxes` function from their `Vizer` library if you check [here](https://github.com/lufficc/Vizer/blob/master/vizer/draw.py) I'd say create your own function to draw the bounding boxes...

@Damon2019 , the code uses GPU by default, but you obviously need a Cuda-enabled system. The configurations are at `ssd/configs/defaults.py` where `_C.MODEL.DEVICE = "cuda"`

@zcf2020DPL , the onnx conversion can be done including the `self.post_processor(detections)` call in the `box_head.py` if the container class is not used in the Postprocessor object in `ssd/modeling/box_head/inference.py`. I have...

Just pass a width parameter with the desired value in the `draw_boxes` function. Like this in the `demo.py` file. ```python 90 drawn_image = draw_boxes(image, boxes, labels, 91 scores, class_names, width=20).astype(np.uint8)...