The homepage is your first point of contact with the Chooch AI Platform.
The “Latest Actions” panel displays your most recent activities in Chooch AI (e.g. creating a dataset or uploading an image).
The “Your Current Plan” panel to the top right displays the active payment plan for your account. Free accounts have limits on their API usage and the models they can create. Free plans cannot share their datasets or models with other users.
The “How-to Videos” and “API Documentation” panels to the right display useful information about using Chooch AI.
The “Edge Devices” panel to the bottom right displays the edge devices you have created.
The “API Key” panel at the bottom displays your account’s secret API key for the Computer Vision API.
The “My Files” page displays your images and videos that have been tagged using Chooch AI’s general detection model.
To upload a file, click on the “Add File” button in the top right. You can upload from your local computer or from your raw data in the Chooch platform (see the “Raw Data” section).
The actions available for any file include:
The “All Folders” panel to the left allows you to organize your files into folders. Clicking on the icon with three dots to the right of a folder’s name allows you to take the same actions (sharing, downloading, renaming, or deleting).
Click on any file in the table to open it and view the Chooch AI model’s output:
The tags are displayed to the right of the image or video in the panel:
You can also manually change the tags generated by the Chooch AI model:
Chooch automatically adds these tags and other metadata to the files you upload. You can view this metadata by downloading the file and viewing it in your computer’s file manager.
The “My Object” page contains your custom-built object detection models.
Unlike image recognition and facial recognition, which are available for inference immediately, Chooch object recognition models require training before use. The training time will depend on the complexity of the model and dataset (i.e. the number of classes and images).
To create an object detection model, you first need to create a dataset:
Once the dataset is created, you then have the option of adding images and videos to the dataset. Click on the dataset name, and then click on “Add new image or video.”
If an image or video lacks any annotations, you will see a red exclamation mark “warning” icon in the bottom right corner:
To add an annotation to an existing file, click on the “Annotate” button, or click on the file itself. The image or video will display in a new window. Click and drag on the screen to create a new annotation:
In order to train a model with Chooch, your images and videos must have at least 30 annotations for each class in the dataset. This requirement applies regardless of the size of the training data (e.g. you can have 1 annotation each for 30 images or 30 annotations for a single image).
Once your dataset has been annotated, click on the “Create Model” button at the top right. Give a name to the model and select the dataset to use from the drop-down menu, then click “Create” to start the training process.
While the model is training, you’ll see a spinning yellow circle in the “Status” column of the table. A green check mark indicates that the model has completed training.
In the “Training” column, you can see a “View Log” link. Click on it to see detailed output from the model’s training, and a chart representing the training process. You can click on the chart to open the image in higher resolution in a new window. The chart represents the number of iterations (i.e. the length of training) plotted against the loss (a rough approximation of the model’s performance). You should see the loss continue to decrease as the number of iterations increases.
The “Mode” column represents a parameter for your AI model:
When training is complete, you can switch between these two modes and check the difference in the results for yourself.
You can test a trained model by uploading some test data to the Chooch AI Platform. The model runs inference on these test images or videos, and then shows the results in a table:
If the model produced no detection at all for an object, click on the “Missing results?” link at the top right. A dialog box will appear asking for the missing object’s class. This test image will then automatically join your dataset, and the AI Platform will allow you to add the annotation for the missing object.
When working with videos, the model will separate the file into one-second chunks, performing inference on each one separately. For a video file, the results panel will display both the object detected and the time in the video at which it was detected:
To view a complete summary of the model’s test results, click on the “Report” button at the top right. This page displays information such as the number of images tested and the model’s precision, recall, and F1 score (three different ways of measuring the accuracy of an AI model). You can also download this report as a Microsoft Excel spreadsheet.
The “My Image” page contains your custom-built image recognition models.
For each of your models, the table displays information such as:
To create a new image recognition model:
Once the images load, Chooch’s pre-trained facial recognition model is available for use. To test its performance, click on “Test via Upload” at the top right, and select images that you have not already uploaded. The model will return its best guess as to the most appropriate image tags.
To edit a class in the model, click on the class name. From there, you can add new images or select images to delete.
The “My Face” page contains your custom-built facial recognition models.
To create a new facial recognition model:
Once the images load, Chooch’s pre-trained facial recognition model is available for use. To test its performance, click on “Test via Upload” at the top right, and select images that you have not already uploaded. The model will return its best guesses as to the person’s identity:
To edit a person in the model, click on the person’s name. From there, you can add new images or select images to delete.
Finally, to share a model, click on the “Share” icon under the Actions column:
The “Public Models” page contains pre-trained AI models that are available for public use.
The four categories of AI models available here are:
For now, object detection and facial detection models can be only deployed to edge AI devices.
Click on one of these categories to view the list of models available to you. Below is just a sampling of the dozens of pre-built Chooch AI models:
The “My Datasets” page offers shortcuts to your datasets for object recognition, image recognition, and facial recognition.
You can quickly access the datasets of a type by clicking on the relevant button or find a particular dataset by searching for its name in the “Latest Dataset” list.
The “Devices” page displays a list of your edge devices.
This panel offers information about your edge devices, including:
The Version tag displays the current software version of Chooch Edge AI. Clicking on the “Ubuntu Setup Guide” link takes you to a PDF document that outlines how to set up a Chooch Edge AI device.
To create a new edge device, click on the “Create Device” button in the top right:
Please visit the Edge Device Setup page for more information.
The “Raw Data” page acts as cloud storage for your files, similar to solutions such as Google Drive and Dropbox.
To upload a new file, click on the “Add” button at the top right.
The “Support” page offers customer support for Chooch users.
From this page, users can view a list of frequently asked questions, watch how-to videos, or find Chooch’s contact information via email and social media.