February 2019 Summaries

The DeveloperWeek SF Hackathon, the largest challenge-driven hackathon in the U.S., took place in San Francisco over the weekend of February 23rd-24th, gathering hundreds of developers, designers, and entrepreneurs to tackle technical challenges with the aid of sponsor-provided technologies. Participants formed teams to create diverse projects, from web and mobile apps to algorithms and data visualizations, with all prizes linked to the effective use of sponsor technologies. Clarifai, an official sponsor, provided open access to its computer vision APIs, which were used in over 30 final applications, including three of the top five finalist projects. The event concluded with rapid project pitches and awards, and Clarifai's engagement allowed them to receive direct user feedback. The hackathon offered ample food and caffeine, and judges evaluated projects based on progress, concept feasibility, and potential for startup viability.

Machine learning has revolutionized the way we search through photos, offering an automated alternative to manual sorting, exemplified by the Clarifai API, which provides image recognition services without requiring machine learning expertise. Users can create applications that allow for visual searches by training models on specific concepts, such as identifying a golden retriever, through a straightforward process involving the Clarifai API. This process includes creating an account, setting up a development environment, uploading labeled images, and training custom models to recognize user-defined concepts. The API supports various languages and allows for both image-based and concept-based searches, providing ranked similarity scores for visual matches. By leveraging Clarifai's pre-trained models and adding custom concepts, users can enhance the accuracy and specificity of their searches, ultimately creating personal visual search applications that learn and improve over time.

Image classification, object detection, and segmentation are key techniques in computer vision that allow machines to interpret and analyze visual data. Image classification involves assigning labels to an entire image or video to identify what is present, such as differentiating between a beach and a pool. Object detection takes this a step further by identifying and localizing specific objects within an image using bounding boxes, which is useful for identifying individual elements like cars on a street. Segmentation goes beyond detection by labeling each pixel in an image, providing detailed outlines of objects, which is particularly beneficial for isolating objects from their backgrounds, such as tagging a shirt in a fashion photo. These techniques collectively enhance computer vision capabilities, enabling machines to process and understand visual information more like humans, thanks to advancements in machine learning and artificial intelligence.

The text is a comprehensive guide on developing an Android alarm clock app that integrates image recognition technology using the Clarifai API to deactivate the alarm only when a specific object is photographed. The process begins by setting up a Clarifai account and obtaining an API key, which facilitates the app's access to Clarifai's computer vision capabilities. The guide details the step-by-step setup in Android Studio, including creating necessary files and configuring the app's manifest to enable camera and internet access. Users set an alarm in the app’s MainActivity by specifying a time and an object, which, when captured by the camera in AlarmActivity, is processed through Clarifai’s general model to verify the object's presence. The app includes a MediaPlayer to play a ringtone and employs an asynchronous task to handle image recognition, ensuring the main thread remains responsive. The guide concludes by suggesting further enhancements, such as repeating alarms, custom sounds, and training Clarifai for personalized object detection, thus encouraging users to explore the potential of image recognition technology in their app development endeavors.

Advances in machine learning have simplified the process of identifying objects in videos, as demonstrated by a tutorial on using Clarifai's APIs to analyze video clips. By utilizing Python and tools such as PyCharm, users can set up an environment to access Clarifai's video recognition capabilities, allowing them to receive JSON responses with detailed predictions. The tutorial guides users through setting up the necessary Python packages, obtaining an API key, and understanding how to send videos for analysis. The Clarifai API returns predictions with confidence scores, and the tutorial provides code to filter and display tags with high confidence for each frame in the video. The approach is versatile, enabling users to incorporate video object recognition into applications, with potential real-time implementation using Clarifai's SDKs.

Adopting a computer vision AI platform involves choosing between three main options: Cloud API, On-Premises, and Software Development Kit (SDK), each offering distinct benefits to businesses. Cloud APIs allow for the use of curated data sets from AI providers, offering improved model accuracy and requiring no maintenance, while ensuring data protection. On-Premises solutions keep data within a private network, providing greater control for large companies with sensitive information, although they lack the collective learning advantage of cloud-based solutions. SDKs enable app development for edge devices, processing data locally without needing an internet connection, which is beneficial in remote areas, but they typically offer reduced functionality compared to Cloud API or On-Premises options. Each of these methods allows users to harness the capabilities of computer vision, whether through internet connectivity, cloud access, or mobile devices.

The POPSUGAR Twinning app, which matches users' selfies with their celebrity lookalikes, gained widespread popularity after Ellen DeGeneres featured it on her talk show and social media. The app uses Clarifai's pre-built celebrity model, allowing developers to launch it quickly without extensive resources. This model facilitates easy implementation through API calls and is regularly updated and maintained. While virality is unpredictable, having scalable infrastructure is crucial for handling sudden spikes in usage, often necessitating third-party providers like AWS for reliable uptime and maintenance. POPSUGAR benefited from Clarifai's support, which allowed them to rapidly scale without server issues, demonstrating the advantages of using external services for app development.