Real_time_object_detection_using_tensorflow/MR_Object_Detection.txt at master · MRobalinho/Real_time_object_detection_using_tensorflow · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
Real-time and video processing object detection using Tensorflow, OpenCV and Docker.

https://towardsdatascience.com/real-time-and-video-processing-object-detection-using-tensorflow-opencv-and-docker-2be1694726e5

1. Clone Git repository
https://github.com/lbeaucourt/Object-detection

2.Install Docker
A plataform for developers that are building containerized applications.
It is a native Windows application that provides an easy-to-use development environment for building, shipping, and running dockerized apps
https://www.docker.com/products/docker-desktop
https://store.docker.com/editions/community/docker-ce-desktop-windows

3. Build docker image:
docker build -t realtime-objectdetection .

4. Configure script (see bellow)

Launch script:

bash runDocker.sh

5. Configuring
Configuration is made in exec.sh at python function call:
All possible arguments are:

	-n (--num-frames): type=int, default=0: # of frames to loop over for FPS test

	-d (--display), type=int, default=0: Whether or not frames should be displayed

	-o (--output), type=int, default=0: Whether or not modified videos shall be writen

	-on (--output-name), type=str, default="output": Name of the output video file

	-I (--input-device), type=int, default=0: Device number input

	-i (--input-videos), type=str, default="": Path to videos input, overwrite device input if used

	-w (--num-workers), type=int, default=2: Number of workers

	-q-size (--queue-size), type=int, default=5: Size of the queue

	-l (--logger-debug), type=int, default=0: Print logger debug

5 Suggested numbers of workers and queues size:

Webcam stream: default values
	** Video stream: 20 workers, 150 queue size (Maybe little hand tunning could be done)
	** Inputs file are in inputs/ folder

Outputs file are in outputs/ folder (.avi)


Example from Tensorflow Video-Detection
https://towardsdatascience.com/building-a-real-time-object-recognition-app-with-tensorflow-and-opencv-b7a2b4ebdc32
https://github.com/tensorflow/models/blob/master/research/object_detection/object_detection_tutorial.ipynb

Research
https://www.pyimagesearch.com/2015/12/21/increasing-webcam-fps-with-python-and-opencv/

Pre-trained models Tensorflow detection model zoo- COCO Dataset
provide a collection of detection models pre-trained on the COCO dataset.
These models can be useful for out-of-the-box inference if you are interested in categories already
in COCO (e.g., humans, cars, etc). They are also useful for initializing your models when training on novel datasets.
https://github.com/tensorflow/models/blob/477ed41e7e4e8a8443bc633846eb01e2182dc68a/object_detection/g3doc/detection_model_zoo.md


Instalar a virtualização
Consultar:  https://docs.microsoft.com/pt-br/virtualization/hyper-v-on-windows/quick-start/enable-hyper-v
No Windows
Powershell (executar como administrador)
>DISM /Online /Enable-Feature /All /FeatureName:Microsoft-Hyper-V