IMAGE RECOGNITION & OBJECT RECOGNITION WITH COMPUTER VISION

Our team of AI experts has deep knowledge of, and experience with, computer vision projects for industry clients—from scoping and planning to development and deployment. We’ve used this expertise to build a powerful yet user-friendly AI solution for the masses: the Chooch AI platform.
The services of the Chooch AI platform include data collection, AI training, model deployment, and inference. By leveraging the Chooch AI platform within your business, you can get split-second, highly accurate results from your images and videos, both in the cloud and on edge devices.

The Chooch AI platform is capable of performing both image recognition and object recognition. What’s the difference between image recognition and object recognition?

• The goal of image recognition is to classify an entire image into a particular category. For example, if we have a dataset that contains images of dogs, an image recognition model might classify these images into categories based on the dog breed depicted. Each image in the dataset on which the model is trained is given a label that denotes its contents. 
• The goal of object recognition (also known as object detection) is to detect the presence of particular objects in an image. For example, if our dataset contained images of two or more dogs, an object recognition model would identify the breed of each dog depicted in the image, as well as its position within the image. Each image in the dataset on which the model is trained is given one or more annotations (either rectangular bounding boxes or a polygonal shape), that contain the label and position of each object in the image.

The clients of Chooch AI have seen tremendous benefits from using our image recognition and object recognition services—so how can you find out if our AI platform is right for you? Keep reading to find out how to try Chooch AI’s image recognition service and object recognition service for yourself.

Chooch AI Image Recognition Demo

It’s easy to get an image recognition demo through the Chooch AI platform: our web-based AI demo gives you an immediate demonstration of the platform’s capabilities. After you upload an image, the pre-trained Chooch AI model will try to match its content with the tags stored within its vast Perception Library. From types of leaves to bacteria under a microscope, the possibilities with the Chooch AI platform are nearly endless.

Chooch AI Object Recognition Demo

As with image recognition, Chooch makes it easy to get an object recognition demo. When you request a live demo, you can use the Chooch AI dashboard to train your own object recognition models using an annotated dataset of images or videos. The Chooch AI platform can provide a highly accurate object detection demo even for smaller datasets.

Chooch AI’s image recognition API is the easiest way to get started bringing the power of AI and computer vision to your organization.

Image Recognition API

Given an input URL that links to an image or video, the image recognition API will return its best guesses as to the contents of that image. Below is a simple example of how to query the image recognition REST API using Python:

import requests
import json
url = ‘https://api.chooch.ai/predict/image?apikey=346g5717-1sd3-35h6-9104-b8h5c819dn19’
files = {‘image’: open(‘local_image.jpg’, ‘rb’)}
response = requests.post(url, files=files)
print(response.content)

In response, the image recognition API will return a JSON output containing the following response fields:

• Url: The URL of the given image file.
• Status: The status of the API request. When successful, the status is “ok”.
• Predictions: The model’s predictions made on the image. Predictions are provided as a list and have class_title and order fields, where class_title is the name of the class predicted, and order is the order of relevancy of the particular class.
• Sub_predictions: Predictions made under dense classification, where the image or frame is segmented into parts and classified based on the segments.
• Text_value: Predictions for text contained in the image (optional).
• Face_name: Predictions for the names of faces recognized in the image (optional).
• Face_count: The number of faces detected in the image (optional).
• Coordinates: The pixel coordinates of an object in the format x1, x2, y1, y2.
• Count: The number of times an object or concept appears in an image or a frame.
• File_ID: A unique ID created for the image to identify it for future reference.

Object Recognition API

As you may have noticed, Chooch AI’s image recognition API performs both image recognition (in the predictions field) and object recognition (in the coordinates field) simultaneously. You can also train your own custom object recognition model and access it through the Chooch AI object recognition API.

Like the image recognition API, querying the Chooch AI custom object recognition API is simple:

import requests import json url = ‘https://api.chooch.ai/predict/object_detection/?url=https://choochdashboard.s3.amazonaws.com/truck.jpg&apikey=346g5717-1sd3-35h6-9104-b8h5c819dn19c&model_id=88’ response = requests.post(url) json_data = json.loads(response.content) print(json_data)

A sample response from the object recognition API is below:

{
“status”: “ok”,
“prediction_type”: “object_detection”,
“predictions”: [
{
“class_title”: “person”,
“model_id”: 88,
“score”: 0.74613,
“coordinates”: {
“xmin”: 0,
“ymin”: 35,
“ymax”: 297,
“xmax”: 179
}
}
]
}

Where class_title is the model’s prediction, score is the model’s confidence, and model_id is the unique ID of the custom object recognition model that you have trained.

For more information, check out our short Computer Vision API videos.