Hello.
I flashed my Jetson Xavier AGX with JetPack 4.5.1 that had Deepstream 5.1 preinstalled.
I obtained the UFF model from my h5 and also generated the corresponding engine file. I can perform inference on the UFF file using the mrcnn sample provided in the TensorRT C/C++ samples.
Now I wish to integrate the model into DeepStream.
For ease with Python, I cloned the https://github.com/NVIDIA-AI-IOT/deepstream_python_apps/blob/master/apps/deepstream-segmentation/ sample made a custom config file:
################################################################################
# Copyright (c) 2021, NVIDIA CORPORATION. All rights reserved.
#
# Permission is hereby granted, free of charge, to any person obtaining a
# copy of this software and associated documentation files (the "Software"),
# to deal in the Software without restriction, including without limitation
# the rights to use, copy, modify, merge, publish, distribute, sublicense,
# and/or sell copies of the Software, and to permit persons to whom the
# Software is furnished to do so, subject to the following conditions:
#
# The above copyright notice and this permission notice shall be included in
# all copies or substantial portions of the Software.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
# THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
# FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
# DEALINGS IN THE SOFTWARE.
################################################################################
# Following properties are mandatory when engine files are not specified:
# int8-calib-file(Only in INT8), model-file-format
# Caffemodel mandatory properties: model-file, proto-file, output-blob-names
# UFF: uff-file, input-dims, uff-input-blob-name, output-blob-names
# ONNX: onnx-file
#
# Mandatory properties for detectors:
# num-detected-classes
#
# Optional properties for detectors:
# cluster-mode(Default=Group Rectangles), interval(Primary mode only, Default=0)
# custom-lib-path,
# parse-bbox-func-name
#
# Mandatory properties for classifiers:
# classifier-threshold, is-classifier
#
# Optional properties for classifiers:
# classifier-async-mode(Secondary mode only, Default=false)
#
# Optional properties in secondary mode:
# operate-on-gie-id(Default=0), operate-on-class-ids(Defaults to all classes),
# input-object-min-width, input-object-min-height, input-object-max-width,
# input-object-max-height
#
# Following properties are always recommended:
# batch-size(Default=1)
#
# Other optional properties:
# net-scale-factor(Default=1), network-mode(Default=0 i.e FP32),
# model-color-format(Default=0 i.e. RGB) model-engine-file, labelfile-path,
# mean-file, gie-unique-id(Default=0), offsets, process-mode (Default=1 i.e. primary),
# custom-lib-path, network-mode(Default=0 i.e FP32)
#
# The values in the config file are overridden by values set through GObject
# properties.
[property]
gpu-id=0
net-scale-factor=0.003921568627451
model-color-format=0
uff-file=/home/virus/Desktop/optimisation/res101-holygrail-ep26.uff
model-engine-file=/home/virus/Desktop/optimisation/res101-holygrail-ep26-fp16.engine
infer-dims=3;1024;1024
uff-input-order=0
uff-input-blob-name=input_image
batch-size=1
## 0=FP32, 1=INT8, 2=FP16 mode
network-mode=2
num-detected-classes=4
interval=0
gie-unique-id=1
network-type=2
output-blob-names=mrcnn_mask/Sigmoid
segmentation-threshold=0.5
#parse-bbox-func-name=NvDsInferParseCustomSSD
#custom-lib-path=nvdsinfer_custom_impl_ssd/libnvdsinfer_custom_impl_ssd.so
#scaling-filter=0
#scaling-compute-hw=0
[class-attrs-all]
pre-cluster-threshold=0.0
roi-top-offset=0
roi-bottom-offset=0
detected-min-w=0
detected-min-h=0
detected-max-w=0
detected-max-h=0
# Optional added by pope
labelfile-path=custom_labels.txt
## Per class configuration
#[class-attrs-2]
#threshold=0.6
#roi-top-offset=20
#roi-bottom-offset=10
#detected-min-w=40
#detected-min-h=40
#detected-max-w=400
#detected-max-h=800
On running
python3 deepstream_segmentation.py custom_config.txt /home/virus/Desktop/optimisation/short.jpg /home/virus/Desktop/results1
I get improper masks, as shown in the attached file:
(a simple grey line…)
On printing the mask data, I found that the inference results don’t look like a mask but rather like a detection.
Mask RCNN has the following inputs and outputs
INFO: [Implicit Engine Info]: layers num: 3
0 INPUT kFLOAT input_image 3x1024x1024 ----------------[Input Layer]
1 OUTPUT kFLOAT mrcnn_detection 100x6-----------------[Bounding Box Detection]
2 OUTPUT kFLOAT mrcnn_mask/Sigmoid 100x4x28x28 ----[Mask]
This is what the mask looks like
[popo_notes][mask]
[[-1 -1 -1 ..., 48 -1 -1]
[-1 -1 49 ..., 7 9 43]
[43 43 6 ..., -1 -1 -1]
[-1 -1 -1 ..., 48 44 44]
[-1 -1 -1 ..., 5 5 5]
[ 5 9 7 ..., 44 -1 -1]]
[popo_notes][mask shape] (6, 1024)
I am not able to fiddle much with the outputs of the pipeline plugins as they are not explained in the repo very well. I suspect that the segmentation sample only expects one output: mask whereas my model provides 2: mask+detection. But I am not very sure about this.