Invalid .MAT file error while code generation for Raspberry Pi

5 views (last 30 days)
Error - Invalid MAT file. MAT file should contain a single instance of either a SeriesNetwork, DAGNetwork, yolov2ObjectDetector or ssdObjectDetector object.
I want to deploy a customised deep learning model on Raspberry Pi 4. I have made my own deep learning model and saved it in .MAT format. When I deploy the code I am getting this error but when I load saved model in workspace it shows it is a 1*1 SeriesNetwork. I have loaded model in program as "coder.loadDeepLearningNetwork('smart.mat')". Also in the given webinar of Deep Learning with Raspberry Pi and MATLAB ( ), they have used a pretrained available model (resnet50) in MATLAB but I want use a customised model. My ARM Compute library version is 20.02.1.
Paras Dhekale
Paras Dhekale on 2 Aug 2021
Edited: Paras Dhekale on 2 Aug 2021
Sir, could you elaborate on how to save only one SeriesNetwork object into the MAT file?
Here is my snippet of saving network after training:-
net = trainNetwork(Images,layers,opt);
customnet = net;
save ('customnet.mat');
When I load this custom network in workspace, its value is 1*1 SeriesNetwork and the variables along with values of this network are Layers(27*1 layer), InputNames(1*1 cell) & OutputNames(1*1 cell).

Sign in to comment.

Answers (1)

Sehairi K.
Sehairi K. on 6 Oct 2021
Usually it happens when Matlab detects two different deep neural networks for example DAGNetwork and yolov2ObjectDetector in the same saved mat file, this happens when use a backend network trying to construct an object detetor network. Try to delete the DAGNetwork and save your 'detetctor' again (save the detector not the mat file)

Sign in to comment.

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by