
Ultimate access to all questions.
As an ML engineer at a leading social media company, you're tasked with developing a visual filter for profile photos that accurately detects human faces with bounding boxes. This filter will be integrated into an iOS mobile application. Your primary objectives are to minimize code development time, ensure the model is optimized for mobile inference, and maintain high accuracy in diverse lighting conditions and angles. Given these requirements, what is the best approach? Choose the most suitable option.
A
Develop a custom TensorFlow model from scratch, focusing on lightweight architectures suitable for mobile devices, and convert it to TensorFlow Lite (TFLite) for deployment.
B
Utilize AutoML Vision to train a custom model with your dataset, selecting the 'export for Coral' option to ensure compatibility with edge devices.
C
Use AutoML Vision to train a custom model tailored to your needs, and select the 'export for Core ML' option for seamless integration into iOS applications.
D
Train a model using AutoML Vision and choose the 'export for TensorFlow.js' option to enable deployment in web-based applications alongside the iOS app.