In multidetect merging boxes fails and produces an error.

[BerendWijers]

Hi Dr. Coffey,

I'm assisting a Postdoc and her Master Student in the setup of an analysis pipeline, using Deep Squeak. I've run into a technical issue and I'm hoping you can help me resolve it. I've provided a description below. Thank you in advance!

Describe the bug When using Multi-detect, after the first audio file has been analysed the merging boxes step fails and produces an error causing the GUI to hang on 'Merging boxes..'

The error message is:

Unrecognized function or variable 'h'.

Error in merge_boxes (line 45)
if ~any(Accepted); close(h); return; end

Error in SqueakDetect (line 143)
Calls = merge_boxes(AllBoxes, AllScores, AllClass, audio_info, 1, score_cuttoff, 0);

Error in multinetdect_Callback (line 80)
        Calls = [Calls;
        SqueakDetect(AudioFile,NeuralNetwork,handles.audiofiles(CurrentAudioFile).name,Settings(:,k),j,length(audioselections),networkname)];

Error in gui_mainfcn (line 95)
        feval(varargin{:});

Error in DeepSqueak (line 29)
    gui_mainfcn(gui_State, varargin{:});

Error in matlab.graphics.internal.figfile.FigFile/read>@(hObject,eventdata)DeepSqueak('multinetdect_Callback',hObject,eventdata,guidata(hObject),0)

Error while evaluating UIControl Callback.

To Reproduce Steps to reproduce the behavior: With the attached audio/detection files:

1. In the main windows go to 'multidetect'
2. Select files 1-4 (attached)
3. Set settings for both networks:
   • Networks: Long Rat Detector Yolo R1 and Rat Detector Yolo R1
   • Total Analysis Length: 0
   • Frequency Cut-off High: 90
   • Frequency Cut-off Low: 18
   • Score Threshold: 0.7
   • Append data to filename: 1
4. Wait for the first audio file to be processed
5. See error in MATLAB console

Expected behavior I expect merging boxes step to resolve and the detection to continue with the next file(s).

Additional context Before this error, a warning is issued:

During the 'Loading network' step, this warning is issued.

Warning: While loading an object of class 'nnet.cnn.TrainingOptionsSGDM':
The value of 'CheckpointPath' is invalid. Checkpoint path must be a char array and an existing directory
with write permission. 
> In multinetdect_Callback (line 77)
In gui_mainfcn (line 95)
In DeepSqueak (line 29)
In matlab.graphics.internal.figfile.FigFile/read>@(hObject,eventdata)DeepSqueak('multinetdect_Callback',hObject,eventdata,guidata(hObject),0) 

Software:

• Windows Server 2019
• Matlab R2021a, Update 7
  • MATLAB support for MinGW-w64 C/C++ compiler
  • GPU Coder interface for deep learning libraries
  • Statistics and Machine Learning Toolbox
  • Signal Processing Toolbox
  • Parallel Computing Toolbox
  • Image Processing Toolbox
  • Deep Learning Toolbox
  • Curve Fitting Toolbox
  • Computer Vision Toolbox
• DeepSqueak v3.04
• NVIDIA driver version 516.31

Hardware:

• Virtual Machine
  • CPU: Intel Xeon Gold 6342 2.80 GHz, 11 cores
  • RAM: 88 GB
  • GPU: NVIDIA A10 Tensor Core GPU

Files https://filesender.surf.nl/?s=download&token=721de2d1-e231-4ee9-988f-7d372cc77f73

[DrCoffey]

Hey @BerendWijers

This crash happened because the long network found a few calls but none of them passed the 0.7 score threshold, and DeepSqueak got confused. I fixed the bug, and it should run fine now if you update DeepSqueak.