Skip to content

Latest commit

 

History

History
209 lines (144 loc) · 11.2 KB

detectnet-console-2.md

File metadata and controls

209 lines (144 loc) · 11.2 KB
Raw

Back | Next | Contents
Object Detection

Locating Objects with DetectNet

The previous recognition examples output class probabilities representing the entire input image. Next we're going to focus on object detection, and finding where in the frame various objects are located by extracting their bounding boxes. Unlike image classification, object detection networks are capable of detecting many different objects per frame.

The detectNet object accepts an image as input, and outputs a list of coordinates of the detected bounding boxes along with their classes and confidence values. detectNet is available to use from Python and C++. See below for various pre-trained detection models available for download. The default model used is a 91-class SSD-Mobilenet-v2 model trained on the MS COCO dataset, which achieves realtime inferencing performance on Jetson with TensorRT.

As examples of using the detectNet class, we provide sample programs for C++ and Python:

These samples are able to detect objects in images, videos, and camera feeds. For more info about the various types of input/output streams supported, see the Camera Streaming and Multimedia page.

Detecting Objects from Images

First, let's try using the detectnet program to locates objects in static images. In addition to the input/output paths, there are some additional command-line options:

  • optional --network flag which changes the detection model being used (the default is SSD-Mobilenet-v2).
  • optional --overlay flag which can be comma-separated combinations of box, labels, conf, and none
    • The default is --overlay=box,labels,conf which displays boxes, labels, and confidence values
  • optional --alpha value which sets the alpha blending value used during overlay (the default is 120).
  • optional --threshold value which sets the minimum threshold for detection (the default is 0.5).

If you're using the Docker container, it's recommended to save the output images to the images/test mounted directory. These images will then be easily viewable from your host device under jetson-inference/data/images/test (for more info, see Mounted Data Volumes).

Here are some examples of detecting pedestrians in images with the default SSD-Mobilenet-v2 model:

# C++
$ ./detectnet --network=ssd-mobilenet-v2 images/peds_0.jpg images/test/output.jpg     # --network flag is optional

# Python
$ ./detectnet.py --network=ssd-mobilenet-v2 images/peds_0.jpg images/test/output.jpg  # --network flag is optional

# C++
$ ./detectnet images/peds_1.jpg images/test/output.jpg

# Python
$ ./detectnet.py images/peds_1.jpg images/test/output.jpg

note: the first time you run each model, TensorRT will take a few minutes to optimize the network.
          this optimized network file is then cached to disk, so future runs using the model will load faster.

Below are more detection examples output from the console programs. The 91-class MS COCO dataset that the SSD-based models were trained on include people, vehicles, animals, and assorted types of household objects to detect.

Various images are found under images/ for testing, such as cat_*.jpg, dog_*.jpg, horse_*.jpg, peds_*.jpg, ect.

Processing a Directory or Sequence of Images

If you have multiple images that you'd like to process at one time, you can launch the detectnet program with the path to a directory that contains images or a wildcard sequence:

# C++
./detectnet "images/peds_*.jpg" images/test/peds_output_%i.jpg

# Python
./detectnet.py "images/peds_*.jpg" images/test/peds_output_%i.jpg

note: when using wildcards, always enclose it in quotes ("*.jpg"). Otherwise, the OS will auto-expand the sequence and modify the order of arguments on the command-line, which may result in one of the input images being overwritten by the output.

For more info about loading/saving sequences of images, see the Camera Streaming and Multimedia page.

Processing Video Files

You can also process videos from disk. There are some test videos found on your Jetson under /usr/share/visionworks/sources/data

# C++
./detectnet /usr/share/visionworks/sources/data/pedestrians.mp4 images/test/pedestrians_ssd.mp4

# Python
./detectnet.py /usr/share/visionworks/sources/data/pedestrians.mp4 images/test/pedestrians_ssd.mp4

# C++
./detectnet /usr/share/visionworks/sources/data/parking.avi images/test/parking_ssd.avi

# Python
./detectnet.py /usr/share/visionworks/sources/data/parking.avi images/test/parking_ssd.avi

Remember that you can use the --threshold setting to change the detection sensitivity up or down (the default is 0.5).

Pre-trained Detection Models Available

Below is a table of the pre-trained object detection networks available for download, and the associated --network argument to detectnet used for loading the pre-trained models:

Model CLI argument NetworkType enum Object classes
SSD-Mobilenet-v1 ssd-mobilenet-v1 SSD_MOBILENET_V1 91 (COCO classes)
SSD-Mobilenet-v2 ssd-mobilenet-v2 SSD_MOBILENET_V2 91 (COCO classes)
SSD-Inception-v2 ssd-inception-v2 SSD_INCEPTION_V2 91 (COCO classes)
DetectNet-COCO-Dog coco-dog COCO_DOG dogs
DetectNet-COCO-Bottle coco-bottle COCO_BOTTLE bottles
DetectNet-COCO-Chair coco-chair COCO_CHAIR chairs
DetectNet-COCO-Airplane coco-airplane COCO_AIRPLANE airplanes
ped-100 pednet PEDNET pedestrians
multiped-500 multiped PEDNET_MULTI pedestrians, luggage
facenet-120 facenet FACENET faces

note: to download additional networks, run the Model Downloader tool
             $ cd jetson-inference/tools
             $ ./download-models.sh

Running Different Detection Models

You can specify which model to load by setting the --network flag on the command line to one of the corresponding CLI arguments from the table above. By default, SSD-Mobilenet-v2 if the optional --network flag isn't specified.

For example, if you chose to download SSD-Inception-v2 with the Model Downloader tool, you can use it like so:

# C++
$ ./detectnet --network=ssd-inception-v2 input.jpg output.jpg

# Python
$ ./detectnet.py --network=ssd-inception-v2 input.jpg output.jpg

Source Code

For reference, below is the source code to detectnet.py:

import jetson.inference
import jetson.utils

import argparse
import sys

# parse the command line
parser = argparse.ArgumentParser(description="Locate objects in a live camera stream using an object detection DNN.")

parser.add_argument("input_URI", type=str, default="", nargs='?', help="URI of the input stream")
parser.add_argument("output_URI", type=str, default="", nargs='?', help="URI of the output stream")
parser.add_argument("--network", type=str, default="ssd-mobilenet-v2", help="pre-trained model to load (see below for options)")
parser.add_argument("--overlay", type=str, default="box,labels,conf", help="detection overlay flags (e.g. --overlay=box,labels,conf)\nvalid combinations are:  'box', 'labels', 'conf', 'none'")
parser.add_argument("--threshold", type=float, default=0.5, help="minimum detection threshold to use") 

try:
	opt = parser.parse_known_args()[0]
except:
	print("")
	parser.print_help()
	sys.exit(0)

# load the object detection network
net = jetson.inference.detectNet(opt.network, sys.argv, opt.threshold)

# create video sources & outputs
input = jetson.utils.videoSource(opt.input_URI, argv=sys.argv)
output = jetson.utils.videoOutput(opt.output_URI, argv=sys.argv)

# process frames until the user exits
while True:
	# capture the next image
	img = input.Capture()

	# detect objects in the image (with overlay)
	detections = net.Detect(img, overlay=opt.overlay)

	# print the detections
	print("detected {:d} objects in image".format(len(detections)))

	for detection in detections:
		print(detection)

	# render the image
	output.Render(img)

	# update the title bar
	output.SetStatus("{:s} | Network {:.0f} FPS".format(opt.network, net.GetNetworkFPS()))

	# print out performance info
	net.PrintProfilerTimes()

	# exit on input/output EOS
	if not input.IsStreaming() or not output.IsStreaming():
		break

Next, we'll run object detection on a live camera stream.

Next | Running the Live Camera Detection Demo
Back | Running the Live Camera Recognition Demo

© 2016-2019 NVIDIA | Table of Contents