diff --git a/docs/docs/building-applications/1-grabbing-images.md b/docs/docs/building-applications/1-grabbing-images.md index 64b0bd1b..b8a34f60 100644 --- a/docs/docs/building-applications/1-grabbing-images.md +++ b/docs/docs/building-applications/1-grabbing-images.md @@ -1,3 +1,7 @@ +--- +sidebar_position: 1 +--- + # Grabbing Images Groundlight's SDK accepts images in many popular formats, including PIL, OpenCV, and numpy arrays. diff --git a/docs/docs/building-applications/working-with-detectors.md b/docs/docs/building-applications/2-working-with-detectors.md similarity index 99% rename from docs/docs/building-applications/working-with-detectors.md rename to docs/docs/building-applications/2-working-with-detectors.md index 0da85c34..dda90640 100644 --- a/docs/docs/building-applications/working-with-detectors.md +++ b/docs/docs/building-applications/2-working-with-detectors.md @@ -1,5 +1,5 @@ --- -sidebar_position: 3 +sidebar_position: 2 --- # Working with Detectors diff --git a/docs/docs/building-applications/5-managing-confidence.md b/docs/docs/building-applications/3-managing-confidence.md similarity index 98% rename from docs/docs/building-applications/5-managing-confidence.md rename to docs/docs/building-applications/3-managing-confidence.md index 5aa66f68..757ce8bd 100644 --- a/docs/docs/building-applications/5-managing-confidence.md +++ b/docs/docs/building-applications/3-managing-confidence.md @@ -1,3 +1,6 @@ +--- +sidebar_position: 3 +--- # Confidence Levels Groundlight gives you a simple way to control the trade-off of latency against accuracy. The longer you can wait for an answer to your image query, the better accuracy you can get. In particular, if the ML models are unsure of the best response, they will escalate the image query to more intensive analysis with more complex models and real-time human monitors as needed. Your code can easily wait for this delayed response. Either way, these new results are automatically trained into your models so your next queries will get better results faster. diff --git a/docs/docs/building-applications/handling-errors.md b/docs/docs/building-applications/4-handling-errors.md similarity index 99% rename from docs/docs/building-applications/handling-errors.md rename to docs/docs/building-applications/4-handling-errors.md index 32a6ca17..36bf8766 100644 --- a/docs/docs/building-applications/handling-errors.md +++ b/docs/docs/building-applications/4-handling-errors.md @@ -1,3 +1,7 @@ +--- +sidebar_position: 4 +--- + # Handling Server Errors When building applications with the Groundlight SDK, you may encounter server errors during API calls. This page covers how to handle such errors and build robust code that can gracefully handle exceptions. diff --git a/docs/docs/building-applications/5-async-queries.md b/docs/docs/building-applications/5-async-queries.md new file mode 100644 index 00000000..d14704c2 --- /dev/null +++ b/docs/docs/building-applications/5-async-queries.md @@ -0,0 +1,70 @@ +--- +sidebar_position: 5 +--- + +# Asynchronous Queries + +Groundlight provides a simple interface for submitting asynchronous queries. This is useful for times in which the thread or process or machine submitting image queries is not the same thread or machine that will be retrieving and using the results. For example, you might have a forward deployed robot or camera that submits image queries to Groundlight, and a separate server that retrieves the results and takes action based on them. We will refer to these two machines as the **submitting machine** and the **retrieving machine**. + +## Setup Submitting Machine +On the **submitting machine**, you will need to install the Groundlight Python SDK. Then you can submit image queries asynchronously using the `ask_async` interface (read the full documentation [here](pathname:///python-sdk/api-reference-docs/#groundlight.client.Groundlight.ask_async)). `ask_async` submits your query and returns as soon as the query is submitted. It does not wait for an answer to be available prior to returning to minimize the time your program spends interacting with Groundlight. As a result, the `ImageQuery` object `ask_async` returns lacks a `result` (the `result` field will be `None`). This is acceptable for this use case as the **submitting machine** is not interested in the result. Instead, the **submitting machine** just needs to communicate the `ImageQuery.id`s to the **retrieving machine** - this might be done via a database, a message queue, or some other mechanism. For this example, we assume you are using a database where you save the `ImageQuery.id` to it via `db.save(image_query.id)`. + +```python notest +from groundlight import Groundlight +import cv2 +from time import sleep + +detector = gl.get_or_create_detector(name="your_detector_name", query="your_query") + +cam = cv2.VideoCapture(0) # Initialize camera (0 is the default index) + +while True: + _, image = cam.read() # Capture one frame from the camera + image_query = gl.ask_async(detector=detector, image=image) # Submit the frame to Groundlight + db.save(image_query.id) # Save the image_query.id to a database for the retrieving machine to use + sleep(10) # Sleep for 10 seconds before submitting the next query + +cam.release() # Release the camera + +``` + +## Setup Retrieving Machine +On the **retrieving machine** you will need to install the Groundlight Python SDK. Then you can retrieve the results of the image queries submitted by another machine using `get_image_query`. The **retrieving machine** can then use the `ImageQuery.result` to take action based on the result for whatever application you are building. For this example, we assume your application looks up the next image query to process from a database via `db.get_next_image_query_id()` and that this function returns `None` once all `ImageQuery`s are processed. +```python notest +from groundlight import Groundlight + +detector = gl.get_or_create_detector(name="your_detector_name", query="your_query") + +image_query_id = db.get_next_image_query_id() + +while image_query_id is not None: + image_query = gl.get_image_query(id=image_query_id) # retrieve the image query from Groundlight + result = image_query.result + + # take action based on the result of the image query + if result.label == 'YES': + pass # TODO: do something based on your application + elif result.label == 'NO': + pass # TODO: do something based on your application + elif result.label == 'UNCLEAR': + pass # TODO: do something based on your application + + # update image_query_id for next iteration of the loop + image_query_id = db.get_next_image_query_id() +``` + +## Important Considerations +When you submit an image query asynchronously, ML prediction on your query is **not** instant. So attempting to retrieve the result immediately after submitting an async query will likely result in an `UNCLEAR` result as Groundlight is still processing your query. Instead, if your code needs a `result` synchronously we recommend using one of our methods with a polling mechanism to retrieve the result. You can see all of the interfaces available in the documentation [here](pathname:///python-sdk/api-reference-docs/#groundlight.client.Groundlight). + +```python notest +from groundlight import Groundlight +from PIL import Image + +detector = gl.get_or_create_detector(name="your_detector_name", query="your_query") +image = Image.open("/path/to/your/image.jpg") +image_query = gl.ask_async(detector=detector, image=image) # Submit async query to Groundlight +result = image_query.result # This will always be 'None' as you asked asynchronously + +image_query = gl.get_image_query(id=image_query.id) # Immediately retrieve the image query from Groundlight +result = image_query.result # This will likely be 'UNCLEAR' as Groundlight is still processing your query +``` \ No newline at end of file diff --git a/docs/docs/building-applications/edge.md b/docs/docs/building-applications/6-edge.md similarity index 96% rename from docs/docs/building-applications/edge.md rename to docs/docs/building-applications/6-edge.md index 3098d0e0..03f6a21f 100644 --- a/docs/docs/building-applications/edge.md +++ b/docs/docs/building-applications/6-edge.md @@ -1,4 +1,8 @@ -# Using Groundlight on the edge +--- +sidebar_position: 6 +--- + +# Using Groundlight on the Edge If your account has access to edge models, you can download and install them to your edge devices. This allows you to run your model evaluations on the edge, reducing latency, cost, network bandwidth, and energy. diff --git a/docs/docs/building-applications/industrial.md b/docs/docs/building-applications/7-industrial.md similarity index 98% rename from docs/docs/building-applications/industrial.md rename to docs/docs/building-applications/7-industrial.md index 48dd0eef..cf50d3ed 100644 --- a/docs/docs/building-applications/industrial.md +++ b/docs/docs/building-applications/7-industrial.md @@ -1,3 +1,7 @@ +--- +sidebar_position: 7 +--- + # Industrial and Manufacturing Applications Modern natural language-based computer vision is transforming industrial and manufacturing applications by enabling more intuitive interaction with automation systems. Groundlight offers cutting-edge computer vision technology that can be seamlessly integrated into various industrial processes, enhancing efficiency, productivity, and quality control. diff --git a/docs/docs/building-applications/building-applications.md b/docs/docs/building-applications/building-applications.md index 68a5b877..d52b039d 100644 --- a/docs/docs/building-applications/building-applications.md +++ b/docs/docs/building-applications/building-applications.md @@ -37,11 +37,14 @@ Groundlight can be used to [apply modern natural-language-based computer vision ## Further Reading For more in-depth guides on various aspects of building applications with Groundlight, check out the following pages: - -- [Working with Detectors](working-with-detectors.md): Learn how to create, configure, and use detectors in your Groundlight-powered applications. -- [Using Groundlight on the edge](edge.md): Discover how to deploy Groundlight in edge computing environments for improved performance and reduced latency. -- [Handling HTTP errors](handling-errors.md): Understand how to handle and troubleshoot HTTP errors that may occur while using Groundlight. - +- **[Grabbing images](1-grabbing-images.md)**: Understand the intricacies of how to submit images from various input sources to Groundlight. +- **[Working with detectors](2-working-with-detectors.md)**: Learn how to create, configure, and use detectors in your Groundlight-powered applications. +- **[Confidence levels](3-managing-confidence.md)**: Master how to control the trade-off of latency against accuracy by configuring the desired confidence level for your detectors. +- **[Handling server errors](4-handling-errors.md)**: Understand how to handle and troubleshoot HTTP errors that may occur while using Groundlight. +- **[Asynchronous queries](5-async-queries.md)**: Groundlight makes it easy to submit asynchronous queries. Learn how to submit queries asynchronously and retrieve the results later. +- **[Using Groundlight on the edge](6-edge.md)**: Discover how to deploy Groundlight in edge computing environments for improved performance and reduced latency. +- **[Industrial applications](7-industrial.md)**: Learn how to apply modern natural-language-based computer vision to your industrial and manufacturing applications. + By exploring these resources and sample applications, you'll be well on your way to building powerful visual applications using Groundlight's computer vision and natural language capabilities. diff --git a/docs/docs/getting-started/getting-started.mdx b/docs/docs/getting-started/getting-started.mdx index c3a5f716..2fd6d2bd 100644 --- a/docs/docs/getting-started/getting-started.mdx +++ b/docs/docs/getting-started/getting-started.mdx @@ -18,7 +18,7 @@ _Note: The SDK is currently in "beta" phase. Interfaces are subject to change in ### How does it work? -Your images are first analyzed by machine learning (ML) models which are automatically trained on your data. If those models have high enough [confidence](docs/building-applications/managing-confidence), that's your answer. But if the models are unsure, then the images are progressively escalated to more resource-intensive analysis methods up to real-time human review. So what you get is a computer vision system that starts working right away without even needing to first gather and label a dataset. At first it will operate with high latency, because people need to review the image queries. But over time, the ML systems will learn and improve so queries come back faster with higher confidence. +Your images are first analyzed by machine learning (ML) models which are automatically trained on your data. If those models have high enough [confidence](docs/building-applications/3-managing-confidence.md), that's your answer. But if the models are unsure, then the images are progressively escalated to more resource-intensive analysis methods up to real-time human review. So what you get is a computer vision system that starts working right away without even needing to first gather and label a dataset. At first it will operate with high latency, because people need to review the image queries. But over time, the ML systems will learn and improve so queries come back faster with higher confidence. ### Escalation Technology