Comments (18)
dhruvsheth-ai
commented on July 19, 2024
2
Could you try changing -
From 544x320->672x384 and give it a shot again?
Thanks.
from depthai-experiments.
viyer-research
commented on July 19, 2024
1
Hi ***@***.***
I have created a repo and all the test files including the models are
available to recreate the error.
https://github.com/viyer-research/fastreid.git
Let me know if you have any questions.
Thanks,
…On Wed, May 26, 2021 at 6:43 AM Vasanth Iyer ***@***.***> wrote:
Hi ***@***.***
So I am sharing my private folder (MyCLoud) for the scripts and the
models.: Please click the link to download the file from the cloud folder
(I will try to put it into Github soon).
https://files.mycloud.com/home.php?brand=webfiles&seuuid=c14a33e625bccd510818bb235288a3fb&name=DepthAI
The model blobs and the above python script can be downloaded from my
private shared folder (depthai)
On Wed, May 26, 2021 at 4:12 AM Erik ***@***.***> wrote:
> @viyer-research <https://github.com/viyer-research> I am guessing you
> were using video input when you got the error? And could you please share
> github gist instead of copying the code, since the visualization/formating
> is off here. Thanks:)
>
> —
> You are receiving this because you were mentioned.
> Reply to this email directly, view it on GitHub
> <#133 (comment)>,
> or unsubscribe
> <https://github.com/notifications/unsubscribe-auth/AB6233X75ENAYGOJPIOYIATTPS3QPANCNFSM45LEECIA>
> .
>
from depthai-experiments.
Erol444
commented on July 19, 2024
1
That's wonderful to hear! So my suggestion would be to use local OpenVINO's model_downloader, download the model (I believe it will be in onnx format) and then use OpenVINO's model optimizer to convert that into IR (xml/bin) :)
from depthai-experiments.
viyer-research
commented on July 19, 2024
It works now so I got further as the re-id algorithm uses two models one
for detection and another for re0id using cosine metric. The detector seems
to be swappable but the re-id blobs only work with pedestrians as I tried
vehicle-reid-001.blob. I got an error
with dai.Device(pipeline) as device:
RuntimeError: Pipeline - 'NeuralNetwork' node with id: 5, isn't compatible
with forced OpenVINO version
Is there any way to make the re-id model compatible with the existing
network as vehicle-detection-adas-0002.blob works? Looking at the network
naming my blind guess would be if I can get vehicle-reid-0002 instead of
-0001. The model converted I am using is shown below.
[image: image.png]
Thanks,
Vasanth
…On Sat, May 22, 2021 at 10:45 PM Dhruv Sheth ***@***.***> wrote:
Could you try changing -
https://github.com/luxonis/depthai-experiments/blob/f6e909b5235241b875481cdcd78ca9b3333dc3fa/gen2-pedestrian-reidentification/main.py#L54
From 544x320->672x384 and give it a shot again?
Thanks.
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
<#133 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AB6233TDNXRXWSGUX4OFHO3TPB255ANCNFSM45LEECIA>
.
from depthai-experiments.
Erol444
commented on July 19, 2024
Hello @viyer-research, it's possible but you have to compile all models with the same openVino version. And you can specify the openVino version with pipeline.setOpenVINOVersion(depthai.OpenVINO.Version.VERSION_2021_3) (for openVino 2021.3)
from depthai-experiments.
viyer-research
commented on July 19, 2024
Hi,
Thanks, I made it further to convert ALL models to 2020_4 but I am getting
this error:
[14442C1051AFC3D200] [176.371] [NeuralNetwork(5)] [warning] Number of
inference threads assigned for network is 1, assigning 2 will likely yield
in better performance
[14442C1051AFC3D200] [176.371] [NeuralNetwork(5)] [warning] Network
compiled for 4 shaves, maximum available 16, compiling for 8 shaves likely
will yield in better performance
[14442C1051AFC3D200] [176.371] [DetectionNetwork(0)] [warning] Network
compiled for 4 shaves, maximum available 16, compiling for 8 shaves likely
will yield in better performance
[14442C1051AFC3D200] [176.380] [NeuralNetwork(5)] [warning] The issued
warnings are orientative, based on optimal settings for a single network,
if multiple networks are running in parallel the optimal settings may vary
[14442C1051AFC3D200] [176.380] [DetectionNetwork(0)] [warning] The issued
warnings are orientative, based on optimal settings for a single network,
if multiple networks are running in parallel the optimal settings may vary
[14442C1051AFC3D200] [176.842] [NeuralNetwork(5)] [error] Input tensor
'input' (0) exceeds available data range. Data size (13824B), tensor offset
(0), size (129792B) - skipping inferenc
-------------------------
My modified code for version consistency is (as my libraries/header don't
have 2021_3, so I am using 2020_4 and converted accordingly
pipeline.setOpenVINOVersion(version = dai.OpenVINO.Version.VERSION_2020_4)
detection_nn.setBlobPath(str(Path("E:/OpenVINO-Re-ID/vehicle-detection-adas-0002-2020-4.blob").resolve().absolute()))
#Works! with sie (672, 384)
reid_nn.setBlobPath(str(Path("E:/OpenVINO-Re-ID/vehicle-reid-0001-2020-4.blob").resolve().absolute()))
# Vehicle RE-IDOn Sun, May 23, 2021 at 1:38 PM Erik
***@***.***> wrote:
Thanks,
Vasanth
… Hello @viyer-research <https://github.com/viyer-research>, it's possible
but you have to compile all models with the same openVino version. And you
can specify the openVino version with
pipeline.setOpenVINOVersion(depthai.OpenVINO.Version.VERSION_2021_3) (for
openVino 2021.3)
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#133 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AB6233QOMS2L6A5R2LKRAOTTPFDTLANCNFSM45LEECIA>
.
from depthai-experiments.
Erol444
commented on July 19, 2024
'input' (0) exceeds available data range. Data size (13824B), tensor offset
(0), size (129792B) - skipping inference
could you share the whole script and what you are trying to accomplish? This happens because wrong data size is sent to the NN node.
from depthai-experiments.
viyer-research
commented on July 19, 2024
After your suggestion, I was able to upgrade to
depthai.OpenVINO.Version.VERSION_2021_3 and compiled all the models to have
the latest version. I am still getting the same error. The model works with
the re-id blob used in the pedestrian example so I check the size and found
the demo size is very small compared to the vehicle re-id model (8BM). Can
it be an OAK-D memory limitation as it seems to work in the OpenVino GPU
environment (not using the OAK-D). Hope you were able to download my script and the models.
Thanks,
Vasanth
…On Sun, May 23, 2021 at 1:38 PM Erik ***@***.***> wrote:
Hello @viyer-research <https://github.com/viyer-research>, it's possible
but you have to compile all models with the same openVino version. And you
can specify the openVino version with
pipeline.setOpenVINOVersion(depthai.OpenVINO.Version.VERSION_2021_3) (for
openVino 2021.3)
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#133 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AB6233QOMS2L6A5R2LKRAOTTPFDTLANCNFSM45LEECIA>
.
from depthai-experiments.
Erol444
commented on July 19, 2024
@viyer-research I am guessing you were using video input when you got the error? And could you please share github gist instead of copying the code, since the visualization/formating is off here. Thanks:)
from depthai-experiments.
viyer-research
commented on July 19, 2024
Hi ***@***.***
So I am sharing my private folder (MyCLoud) for the scripts and the
models.: Please click the link to download the file from the cloud folder
(I will try to put it into Github soon).
https://files.mycloud.com/home.php?brand=webfiles&seuuid=c14a33e625bccd510818bb235288a3fb&name=DepthAI
The model blobs and the above python script can be downloaded from my
private shared folder (depthai)
…On Wed, May 26, 2021 at 4:12 AM Erik ***@***.***> wrote:
@viyer-research <https://github.com/viyer-research> I am guessing you
were using video input when you got the error? And could you please share
github gist instead of copying the code, since the visualization/formating
is off here. Thanks:)
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#133 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AB6233X75ENAYGOJPIOYIATTPS3QPANCNFSM45LEECIA>
.
from depthai-experiments.
Luxonis-Brandon
commented on July 19, 2024
Thanks!
from depthai-experiments.
Erol444
commented on July 19, 2024
Hello @viyer-research , I checked the code, did you get the vehicle-reid-0001 model working in a standalone (just that model)? It seems that's where the problem occurs. Looking at the https://docs.openvinotoolkit.org/latest/omz_models_model_vehicle_reid_0001.html, it seems like the input is 208x208 not 48x96. And you could just use ImgFrame instead of NNData when feeding your image to the model.
Thanks, Erik
from depthai-experiments.
viyer-research
commented on July 19, 2024
I changed the re-id sizes, and it now works. I may need one more pointer on
model conversion: is there a way to convert vehicle-reid-001.blob to
OpenVino vehicle-reid-001.xml vehicle-reid-001.bin as I need to test it
with C++ samples.
Thanks,
…On Fri, May 28, 2021 at 5:11 AM Erik ***@***.***> wrote:
Hello @viyer-research <https://github.com/viyer-research> , I checked the
code, did you get the vehicle-reid-0001 model working in a standalone
(just that model)? It seems that's where the problem occurs. Looking at the
https://docs.openvinotoolkit.org/latest/omz_models_model_vehicle_reid_0001.html,
it seems like the input is 208x208 not 48x96. And you could just use
ImgFrame instead of NNData when feeding your image to the model.
Thanks, Erik
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#133 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AB6233QRKQGGTZF5WLDBYWDTP5T6FANCNFSM45LEECIA>
.
from depthai-experiments.
viyer-research
commented on July 19, 2024
I have all the models set to 2020.4 and so far everything works fine. I
found some issue when using YOLO (all tiny and normal versions) which I
converted to blob using the site luxonis. When loading the models I get an
error:
Creating tracker
Exception in thread Thread-2:
Traceback (most recent call last):
File "C:\Program Files (x86)\Microsoft Visual
Studio\Shared\Python37_64\lib\threading.py", line 926, in _bootstrap_inner
self.run()
File "C:\Program Files (x86)\Microsoft Visual
Studio\Shared\Python37_64\lib\threading.py", line 870, in run
self._target(*self._args, **self._kwargs)
File "yolo-tiny-vehicle-blob-2020-4.py", line 241, in inference_task
nn_data.setLayer("data", to_planar(det_frame, (208, 208)))
File "yolo-tiny-vehicle-blob-2020-4.py", line 50, in to_planar
resized = cv2.resize(arr, shape)
cv2.error: OpenCV(4.5.1)
C:\Users\appveyor\AppData\Local\Temp\1\pip-req-build-oduouqig\opencv\modules\imgproc\src\resize.cpp:4051:
error: (-215:Assertion failed) !ssize.empty() in funct
ion 'cv::resize'
…---
The code is attached for line 50 which is called from the function below
which is size 208x208 for vehicle-reid (works for mobilenetssd and
person-identification)
nn_data.setLayer("data", to_planar(det_frame, (208,
208)))
___________
The code for the YOLO and mobilnetssd are available in github: https://github.com/viyer-research/fastreid.git
Thanks,
On Fri, May 28, 2021 at 2:50 PM Erik ***@***.***> wrote:
That's wonderful to hear! So my suggestion would be to use local
OpenVINO's model_downloader
<https://docs.openvinotoolkit.org/latest/omz_tools_downloader.html>,
download the model (I believe it will be in onnx format) and then use
OpenVINO's model optimizer
<https://docs.openvinotoolkit.org/latest/openvino_docs_MO_DG_Deep_Learning_Model_Optimizer_DevGuide.html>
to convert that into IR (xml/bin) :)
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#133 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AB6233UL6O6Z7JGISBDMH5DTP7XX5ANCNFSM45LEECIA>
.
from depthai-experiments.
Erol444
commented on July 19, 2024
Hello @viyer-research ,
from the error, it looks like you are passing in an empty frame (variable arr) to the cv2.resize() function. You might want to check that out 😄
from depthai-experiments.
viyer-research
commented on July 19, 2024
Hi,
I checked indeed the ndarray was 0 and it happens after the infered_frame
call. The first two detection was fine (see output below) in my test and
the third one has bbox but empty infered_frame? I don't see much of a
difference in the bbox range but the yolo models are 10k bigger then SSD and re-id.
…
-----------------------------------------------
Creating tracker
bbox: [ 802 173 1247 689]
infered_frame-len: <class 'numpy.ndarray'> 688860
infered_frame: [[[236 225 209]
[236 225 209]
[236 225 209]
...
[ 93 142 144]
[101 150 152]
[106 155 157]]
[[236 225 209]
[236 225 209]
[236 225 209]
...
[ 85 134 136]
[ 84 133 135]
[ 83 132 134]]
[[236 225 209]
[236 225 209]
[236 225 209]
...
[ 77 126 128]
[ 66 115 117]
[ 60 109 111]]
...
[[ 64 76 88]
[ 64 76 88]
[ 64 76 88]
...
[ 87 82 88]
[ 87 82 88]
[ 87 82 88]]
[[ 64 76 88]
[ 64 76 88]
[ 64 76 88]
...
[ 87 82 88]
[ 87 82 88]
[ 87 82 88]]
[[ 64 76 88]
[ 64 76 88]
[ 64 76 88]
...
[ 87 82 88]
[ 87 82 88]
[ 87 82 88]]]
bbox: [1275 500 1110 328]
infered_frame-len: <class 'numpy.ndarray'> 0
-------------------------
On Tue, Jun 1, 2021 at 10:31 AM Erik ***@***.***> wrote:
Hello @viyer-research <https://github.com/viyer-research> ,
from the error, it looks like you are passing in an empty frame (variable
arr) to the cv2.resize() function. You might want to check that out ;)
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#133 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AB6233VGEJGLESFKRYBPAE3TQT4NXANCNFSM45LEECIA>
.
from depthai-experiments.
Erol444
commented on July 19, 2024
Hello @viyer-research , we should move to discord (https://discord.gg/EPsZHkg9Nx) for better communication and we can help you resolve your issues there. Since we already resolved the original issue I will close it.
Thanks, Erik
from depthai-experiments.
viyer-research
commented on July 19, 2024
Thanks, I posted the YOLO null issue just now.
…On Tue, Jun 1, 2021 at 11:52 AM Erik ***@***.***> wrote:
Hello @viyer-research <https://github.com/viyer-research> , we should
move to discord (https://discord.gg/EPsZHkg9Nx) for better communication
and we can help you resolve your issues there. Since we already resolved
the original issue I will close it.
Thanks, Erik
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#133 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AB6233WZ6R2BQKTBPIKSPXTTQUF3DANCNFSM45LEECIA>
.
from depthai-experiments.
Related Issues (20)
- Running YOLOv8 model on OAK D Pro Wide camera. I keep getting a not enough memory error. HOT 2
- OAK-D-SR Camera save color image and depth image by record.py HOT 3
- [gen2-yolo/device-decoding] Visualization doesn't show camera footage or detections HOT 2
- To add mesh generation for gen2-pointcloud and Multiple camera HOT 2
- How to visualize when I record data with gen2-record-replay?
- Why pylance is complaining about NNData(), node, and ImageManipConfig() call in script.py from gen2-gaze-estimation project?(Solved) HOT 2
- Head posture estimation module HOT 2
- segmentation with yolov5
- Head pose estimation HOT 2
- How to load external video or webcam feed and process it on device side HOT 1
- New recommendations on head pose estimation model. HOT 1
- Support for YOLOP HOT 2
- Depth information in deep learning model HOT 3
- Camera extrinsic parameter HOT 1
- Issue with installing requirements for RTSP streaming HOT 1
- Deploy yolov8 on OAK-D HOT 6
- Challenges Installing PyGObject for OKA Camera RTSP Streaming HOT 1
- Can we specify the layers you are interested in for a custom nn ? HOT 8
- Yolo v8 on OAK-D camera HOT 1
- gen2-pedestrian-reidentification freezes while person is detected HOT 3
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from depthai-experiments.