Automated Food Recognition and Nutritional Estimation With a Personal Mobile Electronic Device

• US 20160063692A1
• Filed: 12/11/2014
• Published: 03/03/2016
• Est. Priority Date: 09/03/2014
• Status: Active Grant

• Current Assignee: SRI International, Inc.
• Original Assignee: SRI International, Inc.
• Inventors: Divakaran, Ajay, Yu, Qian, Sawhney, Harpreet S., Zhang, Weiyu
• Granted Patent: US 9,734,426 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 18/23213 with fixed number of clusters, e.g. K-means clustering
• CPC Class Codes: G06F 18/2411 based on the proximity to a decision surface, e.g. support vector machines
• CPC Class Codes: G06T 2207/10024 Color image
• CPC Class Codes: G06T 2207/30128 Food products
• CPC Class Codes: G06V 10/462 Salient features, e.g. scale invariant feature transforms [SIFT]
• CPC Class Codes: G06V 10/50 by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis

Abstract

A food recognition assistant system includes technologies to recognize foods and combinations of foods depicted in a digital picture of food. Some embodiments include technologies to estimate portion size and calories, and to estimate nutritional value of the foods. In some embodiments, data identifying recognized foods and related information are generated in an automated fashion without relying on human assistance to identify the foods. In some embodiments, the system includes technologies for achieving automatic food detection and recognition in a real-life setting with a cluttered background, without the images being taken in a controlled lab setting, and without requiring additional user input (such as user-defined bounding boxes). Some embodiments of the system include technologies for personalizing the food classification based on user-specific habits, location and/or other criteria.

Claims

1. A food recognition assistant embodied in one or more non-transitory machine accessible storage media, the food recognition assistant comprising instructions executable by one or more processors to cause a personal mobile electronic device to:

receive a single digital image;

execute one or more feature detection algorithms using at least two different scales of resolution to determine one or more salient regions of the image without relying on a user to specify the one or more salient regions, the one or more salient regions depicting food;

execute a semantic reasoning technique to identify a food present in the image by comparing a characteristic of the one or more salient regions with a characteristic of a known food; and

by an output device of the personal mobile electronic device, output a name of the identified food.

2. The food recognition assistant of claim 1, comprising instructions executable, with the output device, to output a confidence score associated with the identified food.

3. The food recognition assistant of claim 1, wherein the output device comprises a display device, and the executable instructions are to cause the personal mobile electronic device to display, with the display device, the received image and a name of the identified food in relation to the received image.

4. The food recognition assistant of claim 1, comprising instructions executable to algorithmically estimate a portion size of the identified food.

5. The food recognition assistant of claim 4, comprising instructions executable to determine a nutritional fact associated with the identified food based on the estimated portion size, and, by an output device of the personal mobile electronic device, output the nutritional fact associated with the identified food.

6. The food recognition assistant of claim 1, comprising instructions executable to determine the salient region using a plurality of different image processing techniques at multiple scales, the image processing techniques including at least two of:

a histogram of oriented gradients (HoG), an edge orientation histogram, a scale-invariant feature transform descriptors (SIFT), a shape context technique, and a spectral residual method.

7. The food recognition assistant of claim 1, comprising instructions executable to identify the food algorithmically without relying on input from a human user.

8. The food recognition assistant of claim 1, comprising instructions executable to algorithmically classify the one or more salient regions using a food ontology.

9. The food recognition assistant of claim 1, comprising instructions executable to algorithmically identify a plurality of different characteristics of the one or more salient regions, including at least two of:

color, texture, and size.

10. The food recognition assistant of claim 9, comprising instructions executable to semantically label the characteristics of the one or more salient regions by algorithmically applying natural language descriptors to the image for each identified characteristic.

11. The food recognition assistant of claim 10, comprising instructions executable to compare the natural language descriptors to a library of known food descriptors to identify a food in the one or more salient regions.

12. The food recognition assistant of claim 1, comprising instructions executable to identify a plurality of salient regions in the image, execute a similarity algorithm to group adjacent salient regions based on similarity;

and identify food present in the image by comparing a characteristic of a salient region grouping with a characteristic of a known food.

13. A food recognition assistant embodied in one or more non-transitory machine accessible storage media, the food recognition assistant comprising instructions executable by one or more processors to cause a personal mobile electronic device to:

receive a digital image;

execute one or more feature detection algorithms to determine a plurality of salient regions of the image, each salient region being a region of the image that depicts food; and

execute a semantic reasoning technique to identify a food present in at least one of the salient regions, at least partly based on;

(a) comparing a characteristic of the salient region with a characteristic of a known food, and (b) a likelihood of a different food being present in a different one of the salient regions;

wherein at least one of executing the feature detection algorithms and executing the semantic reasoning technique is performed using at least two different scales of resolution with respect to the image; and

by an output device of the personal mobile electronic device, output data indicative of the identified food and the different food.

14. The food recognition assistant of claim 13, comprising instructions executable to cause the personal mobile electronic device to execute the one or more feature detection algorithms without relying on a human user to define the salient regions.

15. The food recognition assistant of claim 13, comprising instructions executable to, by an output device of the personal mobile electronic device, output a confidence score and nutritional facts associated with the identified food and the different food.

16. The food recognition assistant of claim 13, comprising instructions executable to group the salient regions into salient region groupings using an iterative greedy algorithm.

17. The food recognition assistant of claim 13, comprising instructions executable to identify a food in a salient region using a food ontology.

18. The food recognition assistant of claim 13, comprising instructions executable to identify a food in a salient region using stored information indicative of common pairings of foods.

19. The food recognition assistant of claim 13, comprising instructions executable to execute one or more similarity algorithms to group adjacent salient regions into a salient region grouping based on similarity.

20. A personal mobile electronic device comprising:

one or more processors;

an input/output subsystem coupled to the one or more processors;

a camera coupled to the input/output subsystem;

a display device coupled to the input/output subsystem; and

memory comprising instructions executable by the one or more processors to cause the personal mobile electronic device to;

with the camera, take a single picture of a plate of food;

with the input/output subsystem, transmit the picture of the plate of food to a server computing device;

with the input/output subsystem, receive from the server computing device, data identifying a food present in the picture of the plate of food and an algorithmically determined-confidence level associated with the identified food, wherein the data identifying the food present in the picture of the plate of food results from a multi-scale computer vision analysis of the single picture of the plate of food; and

with the display device, display a label indicative of the identified food.

21. The personal mobile electronic device of claim 20, wherein the instructions are executable to, with the display device, display a label indicative of an algorithmically determined nutritional fact relating to the identified food.