StefanWerner I want to deploy a ResNet50 on a Luxonis OAK-D CM4 PoE device. Since I need the outputs of several layers, I define multiple outputs nodes in the onnx graph. Once I run the depthai pipeline, however, I get the issue NeuralNetwork(4) - Out of memory while creating pool for resulting tensor. Number of frames: 8 each with size 358200232B. It works fine when I just use the last layer's output. What can I do to prevent this OOM issue? Thanks for the help!
StefanWerner Okay, I've just fixed the issue. In order to consume less memory, I changed the number of frames in the pool that, I presume, are pre-allocated. Just set, e.g.,network.setNumPoolFrames(1).
