OpenVINO Model Experiment Package

1. Descriptioin

This project provides a set of useful functions to manipulate, analyze, display, and understand the inference result from Intel(R) Distribution of OpenVINO(TM) toolkit.
The library can be called from an independent Python program or from an Jupyter notebook.
The library openvino_model_experiment_package (omep) includes following functions:

• OpenVINO simplified API : model loading, image inferencing, label reading
• Common data processing : normalize, softmax, maxpooling, index sort, BBox NMS, heatmap NMS, Peak detection
• Data visualize : BBox draw, statistics, histogram, heatmap, classification result
• Model specific data parse : classification, SSD, YOLO, Centernet

2. Example of output

Bounding box drawing

import openvino_model_experiment_package as omep
import cv2
model = 'public/mobilenet-ssd/FP16/mobilenet-ssd.xml'
ie, net, exenet, inblobs, outblobs, inshapes, outshapes = omep.load_IR_model(model)
img_orig = cv2.imread('car_1.bmp')
res = omep.infer_ocv_image(exenet, img_orig, inblobs[0])   # Inference
objs = omep.decode_ssd_result(res[outblobs[0]], threshold=0.8)
omep.display_bboxes(objs, img_orig)

Heatmaps (DBFace)

import openvino_model_experiment_package as omep
import cv2
model = 'dbface-vga/dbface.xml'
ie, net, exenet, inblobs, outblobs, inshapes, outshapes = omep.load_IR_model(model)
img_orig = cv2.imread('selfie.jpg')
res = omep.infer_ocv_image(exenet, img_orig, inblobs[0]) # Inference
hm = res[outblobs[0]]
omep.display_heatmap(hm)

Heatmap on an image

import openvino_model_experiment_package as omep
import cv2
model = 'dbface-vga/dbface.xml'
ie, net, exenet, inblobs, outblobs, inshapes, outshapes = omep.load_IR_model(model)
img_orig = cv2.imread('selfie.jpg')
res = omep.infer_ocv_image(exenet, img_orig, inblobs[0]) # Inference
hm = res[outblobs[2]]
omep.display_heatmap(hm, overlay_img=img_orig)

Heatmap with highlight and lowlight marking

import openvino_model_experiment_package as omep
import cv2
model = 'dbface-vga/dbface.xml'
ie, net, exenet, inblobs, outblobs, inshapes, outshapes = omep.load_IR_model(model)
img_orig = cv2.imread('selfie.jpg')
res = omep.infer_ocv_image(exenet, img_orig, inblobs[0]) # Inference
hm = res[outblobs[2]]
omep.display_heatmap(hm, normalize_flg=True, threshold_l=0.02, threshold_h=0.5)

Heatmap with peak markers

import openvino_model_experiment_package as omep
import cv2
model = 'dbface-vga/dbface.xml'
ie, net, exenet, inblobs, outblobs, inshapes, outshapes = omep.load_IR_model(model)
img_orig = cv2.imread('selfie.jpg')
res = omep.infer_ocv_image(exenet, img_orig, inblobs[0]) # Inference
hm = res[outblobs[2]]
omep.display_heatmap(hm, overlay_img=img_orig, normalize_flg=True, draw_peaks=True, peak_threshold=0.2)

Heatmap from human-pose-estimation-0001 model

import openvino_model_experiment_package as omep
import cv2
model = 'intel/human-pose-estimation-0001/FP16/human-pose-estimation-0001.xml'
ie, net, exenet, inblobs, outblobs, inshapes, outshapes = omep.load_IR_model(model)
img_orig = cv2.imread('people.jpg')
res = omep.infer_ocv_image(exenet, img_orig, inblobs[0]) # Inference
omep.display_heatmap(res['Mconv7_stage2_L2'], overlay_img=img_orig, statistics=False)

Histogram

import openvino_model_experiment_package as omep
import cv2
model = 'dbface-vga/dbface.xml'
ie, net, exenet, inblobs, outblobs, inshapes, outshapes = omep.load_IR_model(model)
img_orig = cv2.imread('selfie.jpg')
res = omep.infer_ocv_image(exenet, img_orig, inblobs[0]) # Inference
hm = res[outblobs[2]]
omep.display_histogram(hm, normalize_flg=False)

3. How to use

Place the openvino_model_experiment_package.py to the same directory as the Python project and import it.

import openvino_model_experiment_package as omep

4. API

Refer to the omep-jupyter-test.ipynb (or openvino_model_experiment_package.py) to learn how to use it.

5. Tested environment

• OpenVINO 2020.3 LTS
• Windows 10 1909