gudovskiy / yoloncs

YOLO object detector for Movidius Neural Compute Stick (NCS)

YOLO for Intel/Movidius Neural Compute Stick (NCS)

News

• Camera App is working.
• YOLOv1 Tiny is working.

Protobuf Model files

./prototxt/

Download Pretrained Caffe Models to ./weights/

• YOLO_tiny: https://drive.google.com/file/d/0Bzy9LxvTYIgKNFEzOEdaZ3U0Nms/view?usp=sharing

Compilation

• Compile .prototxt and corresponding .caffemodel (with the same name) to get NCS graph file. For example: "mvNCCompile prototxt/yolo_tiny_deploy.prototxt -w weights/yolo_tiny_deploy.caffemodel -s 12"
• The compiled binary file "graph" has to be in main folder after this step.

Single Image Script

• Run "yolo_example.py" to process a single image. For example: "python3 py_examples/yolo_example.py images/dog.jpg" to get detections as below.

Camera Input Script

• Run "object_detection_app.py" to process a videos from your camera. For example: "python3 py_examples/object_detection_app.py" to get camera detections as below.
• Modify script arguments if needed.
• Press "q" to exit app.

yoloncs's Issues

Where is the TinyYOLO Model/Weights from?

Hey there,

I noticed there is a difference between this repo's TinyYOLO model (yolo_tiny_deploy.prototxt) and the one generated by the code create_yolo_prototxt.py (i.e. the yolo_tiny_deploy.prototxt at xingwangsfu). May I ask where this repo's TinyYOLO model comes from and how can I train it?

Thank you and appreciate your help!

py_examples error

Hi Everyone
I am getting error like below , could you please help me? Camera light is on and I installed OPENCV that was explained in 'http://milq.github.io/install-opencv-ubuntu-debian/' with OPTION 2. All examples work fine but I am still getting error like below. (ValueError: cannot reshape array of size 16922 into shape (7,7,2,4) this line looks like in trouble but I dont know how I can solve.)

[DEBUG/MainProcess] created semlock with handle 140701902925824
[DEBUG/MainProcess] created semlock with handle 140701902921728
[DEBUG/MainProcess] created semlock with handle 140701902917632
[DEBUG/MainProcess] Queue._after_fork()
[DEBUG/MainProcess] created semlock with handle 140701902913536
[DEBUG/MainProcess] created semlock with handle 140701902909440
[DEBUG/MainProcess] created semlock with handle 140701902905344
[DEBUG/MainProcess] Queue._after_fork()
Found stale device, resetting
Device 0 Address: 1.3.3 - VID/PID 03e7:2150
Starting wait for connect with 2000ms timeout
Found Address: 1.3.3 - VID/PID 03e7:2150
Found EP 0x81 : max packet size is 512 bytes
Found EP 0x01 : max packet size is 512 bytes
Found and opened device
Performing bulk write of 865724 bytes...
Successfully sent 865724 bytes of data in 214.364144 ms (3.851478 MB/s)
Boot successful, device address 1.3.3
Found Address: 1.3.3 - VID/PID 03e7:f63b
done
Booted 1.3.3 -> VSC
[DEBUG/MainProcess] created semlock with handle 140701902901248
[DEBUG/MainProcess] created semlock with handle 140701902897152
[DEBUG/MainProcess] created semlock with handle 140701902893056
[DEBUG/MainProcess] created semlock with handle 140701902888960
[DEBUG/MainProcess] added worker
[DEBUG/ForkPoolWorker-1] Queue._after_fork()
[DEBUG/ForkPoolWorker-1] Queue._after_fork()
[INFO/ForkPoolWorker-1] child process calling self.run()
[DEBUG/MainProcess] added worker
[DEBUG/ForkPoolWorker-2] Queue._after_fork()
[DEBUG/ForkPoolWorker-2] Queue._after_fork()
[INFO/ForkPoolWorker-2] child process calling self.run()
[DEBUG/MainProcess] Queue._start_thread()
[DEBUG/MainProcess] doing self._thread.start()
[DEBUG/MainProcess] starting thread to feed data to pipe
[DEBUG/MainProcess] ... done self._thread.start()
/usr/local/lib/python3.5/dist-packages/skimage/transform/_warps.py:84: UserWarning: The default mode, 'constant', will be changed to 'reflect' in skimage 0.15.
warn("The default mode, 'constant', will be changed to 'reflect' in "
[INFO/ForkPoolWorker-1] process shutting down
[DEBUG/ForkPoolWorker-1] running all "atexit" finalizers with priority >= 0
[DEBUG/ForkPoolWorker-1] running the remaining "atexit" finalizers
Process ForkPoolWorker-1:
Traceback (most recent call last):
File "/usr/lib/python3.5/multiprocessing/process.py", line 249, in _bootstrap
self.run()
File "/usr/lib/python3.5/multiprocessing/process.py", line 93, in run
self._target(*self._args, **self._kwargs)
File "/usr/lib/python3.5/multiprocessing/pool.py", line 103, in worker
initializer(*initargs)
File "py_examples/object_detection_app.py", line 115, in worker
results = interpret_output(out.astype(np.float32), frame.shape[1], frame.shape[0])
File "py_examples/object_detection_app.py", line 55, in interpret_output
boxes = (np.reshape(output[1078:],(7,7,2,4)))
File "/usr/local/lib/python3.5/dist-packages/numpy/core/fromnumeric.py", line 232, in reshape
return _wrapfunc(a, 'reshape', newshape, order=order)
File "/usr/local/lib/python3.5/dist-packages/numpy/core/fromnumeric.py", line 57, in _wrapfunc
return getattr(obj, method)(*args, **kwds)
ValueError: cannot reshape array of size 16922 into shape (7,7,2,4)
[INFO/ForkPoolWorker-1] process exiting with exitcode 1
[DEBUG/MainProcess] cleaning up worker 0
[Level 5/MainProcess] finalizer calling with args (19,) and kwargs {}
[DEBUG/MainProcess] added worker
[DEBUG/ForkPoolWorker-3] Queue._after_fork()
[DEBUG/ForkPoolWorker-3] Queue._after_fork()
[INFO/ForkPoolWorker-3] child process calling self.run()

Which .cfg you use？

Hi ,

I want to convert my cfg file to prototxt format.
I use one of this dir's file(http://github.com/pjreddie/darknet/cfg/yolov1).,
then I use xingwangsfu's tools(https://github.com/xingwangsfu/caffe-yolo),but the result is not good.
Can you tell me which cfg you used?

Thanks

Duplication layer Error

Hi there i wanted to convert tiny yolo cfg to yolo prototxt but i am getting this error.
configparser.DuplicateOptionError: While reading from 'tiny-yolo-azface-fddb.cfg' [line 44]: option 'batch_normalize' in section 'convolutional' already exists

Example typo/error

mvNCCompile prototxt/yolo_tiny_deploy.prototxt -w weights/yolo_tiny**_deploy**.caffemodel -s 12

EIther one needs to change the name of the weight file or use this name for downloaded weight file.

mvNCCompile prototxt/yolo_tiny_deploy.prototxt -w weights/yolo_tiny.caffemodel -s 12

load model

How can I load several models simultaneously?

How can I retrain this Yolo model for transfer learning?

Hi I want to use this TinyYolo model but want it only for some other specified categories.
Can you give me some instructions on how can I retrain and compile it?

Labels file

Hi.
Would it be possible to get the list of labels?
best regards

Can run sample for NCSv1.12

pi@raspberrypi:~/YoloV2NCS-1.0 $ python3 detectionExample/Main.py
Traceback (most recent call last):
File "detectionExample/Main.py", line 5, in
from ObjectWrapper import *
File "/home/pi/YoloV2NCS-1.0/detectionExample/ObjectWrapper.py", line 1, in
from libpydetector import YoloDetector
ImportError: /home/pi/YoloV2NCS-1.0/detectionExample/libpydetector.so: wrong ELF class: ELFCLASS64

I compiled it on another box a move the files, created the lync

can you share instructions for ncsdf 1.09 release?

Hi,

I just bought the Movidius stick and installed ncsdk 1.09 release. Seems the structure has changed a lot. Command line for mvNCCompile is different, and i cannot find the directory for ncapi/networks/YoloTiny/ . Could you please provide suggestions that how I can run tiny yolo in ncsdk1.09 please?

object detection multiprocessing graph load fail

The single image test works.
This appears to be a video capture issue.

The object detection is producing a graph.LoadTensor error

[INFO/ForkPoolWorker-1] process shutting down
[DEBUG/ForkPoolWorker-1] running all "atexit" finalizers with priority >= 0
[DEBUG/ForkPoolWorker-1] running the remaining "atexit" finalizers
Process ForkPoolWorker-1:
Traceback (most recent call last):
  File "/usr/lib/python3.5/multiprocessing/process.py", line 249, in _bootstrap
    self.run()
  File "/usr/lib/python3.5/multiprocessing/process.py", line 93, in run
    self._target(*self._args, **self._kwargs)
  File "/usr/lib/python3.5/multiprocessing/pool.py", line 103, in worker
    initializer(*initargs)
  File "py_examples/object_detection_app.py", line 113, in worker
    graph.LoadTensor(resize(frame/255.0,dim,1)[:,:,(2,1,0)].astype(np.float16), 'user object')
TypeError: unsupported operand type(s) for /: 'NoneType' and 'float'
[INFO/ForkPoolWorker-1] process exiting with exitcode 1
[DEBUG/MainProcess] cleaning up worker 0
[Level 5/MainProcess] finalizer calling <built-in function close> with args (18,) and kwargs {}
[DEBUG/MainProcess] added worker
[DEBUG/ForkPoolWorker-3] Queue._after_fork()
[DEBUG/ForkPoolWorker-3] Queue._after_fork()
[INFO/ForkPoolWorker-3] child process calling self.run()

Error running detectionExample

File " ./YoloV2NCS/detectionExample/ObjectWrapper.py", line 101, in Detect
ObjectWrapper.fifoInHandle[i],
NameError: name 'i' is not defined

Can't run example

Hello,
First of all, thank you for the project. I already installed the project and compiled everything. The problem is when I try to run 'python3 py_examples/yolo_example.py images/dog.jpg' I get the following error:

Device 0 Address: 3 - VID/PID 03e7:2150
Starting wait for connect with 2000ms timeout
Found Address: 3 - VID/PID 03e7:2150
Found EP 0x81 : max packet size is 512 bytes
Found EP 0x01 : max packet size is 512 bytes
Found and opened device
Performing bulk write of 865212 bytes...
Successfully sent 865212 bytes of data in 457.894553 ms (1.802010 MB/s)
Boot successful, device address 3
Found Address: 3 - VID/PID 03e7:f63b
done
Booted 3 -> VSC
Traceback (most recent call last):
File "py_examples/yolo_example.py", line 126, in
graph = device.AllocateGraph(blob)
File "/usr/local/lib/python3.5/dist-packages/mvnc/mvncapi.py", line 201, in AllocateGraph
raise Exception(Status(status))
Exception: mvncStatus.ERROR

Any suggestions? I do have the 'graph' file in the main folder. Thanks again!

How are the pretrained weights generated? How do I generate them?

Are the pretrained weights generated by using create_yolo_caffemodel.py?

If I want to train the model on my own what would the process be? Should I first train yolo by darknet and then use create_yolo_caffemodel.py and create_yolo_prototxt.py?

can't load object_detection.py

Hey, I am getting an error loading this.

[DEBUG/MainProcess] added worker
[DEBUG/ForkPoolWorker-2] Queue._after_fork()
[DEBUG/ForkPoolWorker-2] Queue._after_fork()
[INFO/ForkPoolWorker-2] child process calling self.run()
[DEBUG/MainProcess] Queue._start_thread()
[DEBUG/MainProcess] doing self._thread.start()
[DEBUG/MainProcess] starting thread to feed data to pipe
[DEBUG/MainProcess] ... done self._thread.start()
[INFO/ForkPoolWorker-1] process shutting down
[DEBUG/ForkPoolWorker-1] running all "atexit" finalizers with priority >= 0
[DEBUG/ForkPoolWorker-1] running the remaining "atexit" finalizers
Process ForkPoolWorker-1:
Traceback (most recent call last):
  File "/usr/lib/python3.5/multiprocessing/process.py", line 249, in _bootstrap
    self.run()
  File "/usr/lib/python3.5/multiprocessing/process.py", line 93, in run
    self._target(*self._args, **self._kwargs)
  File "/usr/lib/python3.5/multiprocessing/pool.py", line 103, in worker
    initializer(*initargs)
  File "object_detection_app.py", line 113, in worker
    graph.LoadTensor(resize(frame/255.0,dim,1)[:,:,(2,1,0)].astype(np.float16), 'user object')
TypeError: unsupported operand type(s) for /: 'NoneType' and 'float'
[INFO/ForkPoolWorker-1] process exiting with exitcode 1
[DEBUG/MainProcess] cleaning up worker 0
[Level 5/MainProcess] finalizer calling <built-in function close> with args (18,) and kwargs {}
[DEBUG/MainProcess] added worker
[DEBUG/ForkPoolWorker-3] Queue._after_fork()
[DEBUG/ForkPoolWorker-3] Queue._after_fork()
[INFO/ForkPoolWorker-3] child process calling self.run()

Let me know if you can help

Compilation issue

Facing error during compilation. Any idea on this?

Thanks for the help in advance.

mvNCCompile v02.00, Copyright @ Movidius Ltd 2016

Traceback (most recent call last):
File "/usr/local/bin/mvNCCompile", line 118, in
create_graph(args.network, args.inputnode, args.outputnode, args.outfile, args.nshaves, args.inputsize, args.weights)
File "/usr/local/bin/mvNCCompile", line 101, in create_graph
net = parse_caffe(args, myriad_config)
File "/usr/local/bin/ncsdk/Controllers/CaffeParser.py", line 365, in parse_caffe
if layers[0].type == 'Input':
IndexError: list index out of range