Error in Hailo Application Code Examples

General
, , ,
1

In the Hailo Application Code Examples Repo, there is an example of how to use the hailo api to stream video.

On line 67 there is this statement
layer_from_shape: dict = {infer_results[key].shape:key for key in infer_results.keys()}
This throws an error because the data type of infer_results[key] is a list (not a numpy array).

Actually, I dont get what this is supposed to do.
It is going to return a dictionary where they keys are the tensor shape, and the values are the model name (in this case “yolox_s_leaky/yolox_nms_postprocess”). It iterates over all the keys, but there is only one key in infer_results

Given that there is only one key in infer_results, I dont see how the code

endnodes = [infer_results[layer_from_shape[1, 80, 80, 4]],  # stride 8 
                    infer_results[layer_from_shape[1, 80, 80, 1]],  # stride 8 ...

is supposed to work.

Maybe the .hef file has been changes since this code was written (???). I am using the .hef file specified in the get_resource.sh file
(“https://hailo-model-zoo.s3.eu-west-2.amazonaws.com/ModelZoo/Compiled/v2.6.1/yolox_s_leaky.hef”)

Has anyone got this to work?

2

Hi! For me, the example is also broken, but for a different reason:

Traceback (most recent call last):
  File "/local/workspace/Hailo-Application-Code-Examples/runtime/python/streaming/./yolox_stream_inference.py", line 87, in <module>
    hailo_preds = yolox_post_proc.yolo_postprocessing(endnodes)
  File "/local/workspace/hailo_model_zoo/hailo_model_zoo/core/postprocessing/detection/yolo.py", line 139, in yolo_postprocessing
    detection_boxes, detection_scores = tf.numpy_function(
  File "/local/workspace/hailo_virtualenv/lib/python3.10/site-packages/tensorflow/python/util/traceback_utils.py", line 153, in error_handler
    raise e.with_traceback(filtered_tb) from None
  File "/local/workspace/hailo_model_zoo/hailo_model_zoo/core/postprocessing/detection/yolo.py", line 209, in yolo_postprocess_numpy
    pred = net_out.transpose((0, 3, 1, 2))  # now dims are: [N,C,H,W] as in Gluon.
ValueError: axes don't match array

Is it correct to assume that for now, this example is simply broken?

3

Hey @Nils-Oliver, @mgreiner79,

I’ll investigate and update the implementation. For now, please assume it’s not working until I provide an updated version. Unlike our stable RPi examples repository, this repo is still being stabilized and we’re working on adding more examples.

I’ll keep you updated on the progress.