Hi,
I trained a darknet yolov3 model on my own datasets and got a good accuracy. I deployed it on jetson TX2 using deepstream SDK 4.0 yolo-app (following the steps in README file) but I noticed a big accuracy drop (I tried with both FP16 and FP32). why the accuracy drops? what should I do in this case?
ubuntu: 18.04
tensorRT: 5.1.6
Hi,
A possible issue is the different image color format.
Are you using OpenCV for image IO? If yes, the color format of training should be BGR.
Please update the model configure to use the corresponding color format and try it again.
For example: config_infer_primary_yoloV3.txt
[property]
gpu-id=0
net-scale-factor=1
#0=RGB, 1=BGR
[b]model-color-format=0
[/b]...
Thanks.
Hi,
Yes, I am using opneCV for imageIO. I tried with BGR option but there is no remarkable improvement.
Hi AastaLLL,
Here are my configuration files:
################################################################################
# Copyright (c) 2019, 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.
################################################################################
[application]
enable-perf-measurement=1
perf-measurement-interval-sec=5
#gie-kitti-output-dir=streamscl
[tiled-display]
enable=1
rows=1
columns=1
width=3840
height=2160
gpu-id=0
#(0): nvbuf-mem-default - Default memory allocated, specific to particular platform
#(1): nvbuf-mem-cuda-pinned - Allocate Pinned/Host cuda memory, applicable for Tesla
#(2): nvbuf-mem-cuda-device - Allocate Device cuda memory, applicable for Tesla
#(3): nvbuf-mem-cuda-unified - Allocate Unified cuda memory, applicable for Tesla
#(4): nvbuf-mem-surface-array - Allocate Surface Array memory, applicable for Jetson
nvbuf-memory-type=0
[source0]
enable=1
#Type - 1=CameraV4L2 2=URI 3=MultiURI
type=2
#uri=file://../../samples/streams/sample_1080p_h264.mp4
uri=file://../../sources/objectDetector_Yolo/esb2_5m.mp4
num-sources=1
gpu-id=0
# (0): memtype_device - Memory type Device
# (1): memtype_pinned - Memory type Host Pinned
# (2): memtype_unified - Memory type Unified
cudadec-memtype=0
[sink0]
enable=1
#Type - 1=FakeSink 2=EglSink 3=File
type=3
#1=mp4 2=mkv
container=1
#1=h264 2=h265
codec=2
sync=0
#iframeinterval=10
bitrate=2000000
output-file=out.mp4
source-id=0
gpu-id=0
nvbuf-memory-type=0
[osd]
enable=1
gpu-id=0
border-width=1
text-size=15
text-color=1;1;1;1;
text-bg-color=0.3;0.3;0.3;1
font=Serif
show-clock=0
clock-x-offset=800
clock-y-offset=820
clock-text-size=12
clock-color=1;0;0;0
nvbuf-memory-type=0
[streammux]
gpu-id=0
##Boolean property to inform muxer that sources are live
live-source=0
batch-size=1
##time out in usec, to wait after the first buffer is available
##to push the batch even if the complete batch is not formed
batched-push-timeout=40000
## Set muxer output width and height
width=1920
height=1080
##Enable to maintain aspect ratio wrt source, and allow black borders, works
##along with width, height properties
enable-padding=0
nvbuf-memory-type=0
# config-file property is mandatory for any gie section.
# Other properties are optional and if set will override the properties set in
# the infer config file.
[primary-gie]
enable=1
gpu-id=0
#model-engine-file=model_b1_int8.engine
labelfile-path=labels.txt
batch-size=1
#Required by the app for OSD, not a plugin property
bbox-border-color0=1;0;0;1
bbox-border-color1=0;1;1;1
bbox-border-color2=0;0;1;1
bbox-border-color3=0;1;0;1
#interval=0
gie-unique-id=1
nvbuf-memory-type=0
config-file=config_infer_primary_yoloV3.txt
[tests]
file-loop=0
################################################################################
# Copyright (c) 2019, 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:
# enable-dbscan(Default=false), 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, gie-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=1
#0=RGB, 1=BGR
model-color-format=1
custom-network-config=yolov3.cfg
model-file=yolov3.weights
#model-engine-file=model_b1_fp32.engine
labelfile-path=labels.txt
#int8-calib-file=yolov3-calibration.table.trt5.1
## 0=FP32, 1=INT8, 2=FP16 mode
network-mode=0
num-detected-classes=2
gie-unique-id=1
is-classifier=0
maintain-aspect-ratio=1
parse-bbox-func-name=NvDsInferParseCustomYoloV3
custom-lib-path=nvdsinfer_custom_impl_Yolo/libnvdsinfer_custom_impl_Yolo.so
and here the configuration file:
[net]
# Testing
batch=64
subdivisions=64
# Training
# batch=64
# subdivisions=16
width=832
height=832
channels=3
momentum=0.9
decay=0.0005
angle=35
saturation = 1.5
exposure = 1.5
hue=.1
learning_rate=0.001
burn_in=1000
max_batches = 6000
policy=steps
steps=4800,5600
scales=.1,.1
[convolutional]
batch_normalize=1
filters=32
size=3
stride=1
pad=1
activation=leaky
# Downsample
[convolutional]
batch_normalize=1
filters=64
size=3
stride=2
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=32
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=64
size=3
stride=1
pad=1
activation=leaky
[shortcut]
from=-3
activation=linear
# Downsample
[convolutional]
batch_normalize=1
filters=128
size=3
stride=2
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=64
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=128
size=3
stride=1
pad=1
activation=leaky
[shortcut]
from=-3
activation=linear
[convolutional]
batch_normalize=1
filters=64
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=128
size=3
stride=1
pad=1
activation=leaky
[shortcut]
from=-3
activation=linear
# Downsample
[convolutional]
batch_normalize=1
filters=256
size=3
stride=2
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=128
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=256
size=3
stride=1
pad=1
activation=leaky
[shortcut]
from=-3
activation=linear
[convolutional]
batch_normalize=1
filters=128
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=256
size=3
stride=1
pad=1
activation=leaky
[shortcut]
from=-3
activation=linear
[convolutional]
batch_normalize=1
filters=128
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=256
size=3
stride=1
pad=1
activation=leaky
[shortcut]
from=-3
activation=linear
[convolutional]
batch_normalize=1
filters=128
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=256
size=3
stride=1
pad=1
activation=leaky
[shortcut]
from=-3
activation=linear
[convolutional]
batch_normalize=1
filters=128
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=256
size=3
stride=1
pad=1
activation=leaky
[shortcut]
from=-3
activation=linear
[convolutional]
batch_normalize=1
filters=128
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=256
size=3
stride=1
pad=1
activation=leaky
[shortcut]
from=-3
activation=linear
[convolutional]
batch_normalize=1
filters=128
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=256
size=3
stride=1
pad=1
activation=leaky
[shortcut]
from=-3
activation=linear
[convolutional]
batch_normalize=1
filters=128
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=256
size=3
stride=1
pad=1
activation=leaky
[shortcut]
from=-3
activation=linear
# Downsample
[convolutional]
batch_normalize=1
filters=512
size=3
stride=2
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=256
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=512
size=3
stride=1
pad=1
activation=leaky
[shortcut]
from=-3
activation=linear
[convolutional]
batch_normalize=1
filters=256
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=512
size=3
stride=1
pad=1
activation=leaky
[shortcut]
from=-3
activation=linear
[convolutional]
batch_normalize=1
filters=256
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=512
size=3
stride=1
pad=1
activation=leaky
[shortcut]
from=-3
activation=linear
[convolutional]
batch_normalize=1
filters=256
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=512
size=3
stride=1
pad=1
activation=leaky
[shortcut]
from=-3
activation=linear
[convolutional]
batch_normalize=1
filters=256
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=512
size=3
stride=1
pad=1
activation=leaky
[shortcut]
from=-3
activation=linear
[convolutional]
batch_normalize=1
filters=256
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=512
size=3
stride=1
pad=1
activation=leaky
[shortcut]
from=-3
activation=linear
[convolutional]
batch_normalize=1
filters=256
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=512
size=3
stride=1
pad=1
activation=leaky
[shortcut]
from=-3
activation=linear
[convolutional]
batch_normalize=1
filters=256
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=512
size=3
stride=1
pad=1
activation=leaky
[shortcut]
from=-3
activation=linear
# Downsample
[convolutional]
batch_normalize=1
filters=1024
size=3
stride=2
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=512
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=1024
size=3
stride=1
pad=1
activation=leaky
[shortcut]
from=-3
activation=linear
[convolutional]
batch_normalize=1
filters=512
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=1024
size=3
stride=1
pad=1
activation=leaky
[shortcut]
from=-3
activation=linear
[convolutional]
batch_normalize=1
filters=512
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=1024
size=3
stride=1
pad=1
activation=leaky
[shortcut]
from=-3
activation=linear
[convolutional]
batch_normalize=1
filters=512
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=1024
size=3
stride=1
pad=1
activation=leaky
[shortcut]
from=-3
activation=linear
######################
[convolutional]
batch_normalize=1
filters=512
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
size=3
stride=1
pad=1
filters=1024
activation=leaky
[convolutional]
batch_normalize=1
filters=512
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
size=3
stride=1
pad=1
filters=1024
activation=leaky
[convolutional]
batch_normalize=1
filters=512
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
size=3
stride=1
pad=1
filters=1024
activation=leaky
[convolutional]
size=1
stride=1
pad=1
filters=21
activation=linear
[yolo]
mask = 6,7,8
anchors = 23,43, 37,66, 60,53, 43,115, 76,81, 108,107, 79,182, 151,171, 180,303
classes=2
num=9
jitter=.3
ignore_thresh = .7
truth_thresh = 1
random=1
[route]
layers = -4
[convolutional]
batch_normalize=1
filters=256
size=1
stride=1
pad=1
activation=leaky
[upsample]
stride=2
[route]
layers = -1, 61
[convolutional]
batch_normalize=1
filters=256
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
size=3
stride=1
pad=1
filters=512
activation=leaky
[convolutional]
batch_normalize=1
filters=256
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
size=3
stride=1
pad=1
filters=512
activation=leaky
[convolutional]
batch_normalize=1
filters=256
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
size=3
stride=1
pad=1
filters=512
activation=leaky
[convolutional]
size=1
stride=1
pad=1
filters=21
activation=linear
[yolo]
mask = 3,4,5
anchors = 23,43, 37,66, 60,53, 43,115, 76,81, 108,107, 79,182, 151,171, 180,303
classes=2
num=9
jitter=.3
ignore_thresh = .7
truth_thresh = 1
random=1
[route]
layers = -4
[convolutional]
batch_normalize=1
filters=128
size=1
stride=1
pad=1
activation=leaky
[upsample]
stride=2
[route]
layers = -1, 36
[convolutional]
batch_normalize=1
filters=128
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
size=3
stride=1
pad=1
filters=256
activation=leaky
[convolutional]
batch_normalize=1
filters=128
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
size=3
stride=1
pad=1
filters=256
activation=leaky
[convolutional]
batch_normalize=1
filters=128
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
size=3
stride=1
pad=1
filters=256
activation=leaky
[convolutional]
size=1
stride=1
pad=1
filters=21
activation=linear
[yolo]
mask = 0,1,2
anchors = 23,43, 37,66, 60,53, 43,115, 76,81, 108,107, 79,182, 151,171, 180,303
classes=2
num=9
jitter=.3
ignore_thresh = .7
truth_thresh = 1
random=1
Hi,
Could you share your model with us so we can check it further?
Thanks.
Hi AastaLLL,
Thank you for your help, is it possible to give me your email so I can share with you the files. sorry for that but the model weights and inputs are not public.
Thanks,
Hi,
You can pass it through private message on the forum.
Thanks.
Hi,
I have sent you the files in a private message.
Thanks
Hi AastaLLL,
any updates…
Thanks
Hi,
Sorry that we still need some time for this.
Will update more information with you soon.
Thanks
Hi AastaLLL,
Any updates on this issue?
I tried to run it on jetson xavier, but same problem. could it be because Alexey/darkent is different than pjreddie/darknet, maybe different preprocessing ?
Thanks
Hi,
There is a similar topic on the Deepstream board.
Could you check if the solution also works for you?
https://devtalk.nvidia.com/default/topic/1066154/deepstream-sdk/different-result-with-custom-tiny_yolo-model-in-deepstream-vs-cv2-dnn-module/
You will need to update the anchor/mask and parameters, which can be found in this document:
https://docs.nvidia.com/metropolis/deepstream/Custom_YOLO_Model_in_the_DeepStream_YOLO_App.pdf#page=3
Thanks.
Hi,
Thank you for your feedback. I already updated my anchor/masks and parameters in nvdsparsebbox_Yolo.cpp file. but still getting low accuracy.
Thanks
Hi,
Could you share the sample with your update for us debugging?
Thanks.
Hi AastaLLL,
I have updated the files and send them in a private message.
Thanks
Hi,
We found a different in the official YOLOv3 and your customized version.
angle=35
Please noticed that we don’t apply the corresponding handling for this parameter in Deepstream sample.
So the inference will be applied on the image without any rotation.
Do you think the difference is caused by this?
Thanks.