This episode dives into the growing trend of mobile apps that estimate calorie counts from a simple photo of your meal. Lucas and Luna explore the technical backbone: how on-device AI models segment food items, estimate volume, and cross-reference nutritional databases — all without sending your lunch to the cloud. They focus on a specific case: the Open Food Facts dataset, which contains over 3 million products, and how app developers like those behind the app 'Calorie Mama' have used it to build real-time estimators. The hosts discuss the challenge of portion size estimation — where most apps still struggle — and the trade-offs between accuracy and latency. They also touch on how Apple's Core ML and Google's ML Kit enable these features on-device, preserving user privacy. A concrete takeaway: the best current apps achieve around 70-80% accuracy for simple meals but drop to 50% for complex dishes like casseroles. The conversation concludes with a look at how this technology could integrate with dietary tracking and health recommendations, and why the next frontier is real-time video analysis of your plate.

#OnDeviceAI #FoodCalorieEstimation #MobileApps #ComputerVision #CoreML #MLKit #OpenFoodFacts #CalorieMama #NutritionTracking #PortionSizeEstimation #DeepLearning #ImageSegmentation #Privacy #Tech #Business #FexingoBusiness #BusinessPodcast #MobileDevelopment

Keep every episode free: buymeacoffee.com/fexingo

Lucas and Luna explore how modern apps analyse piano chords on-device using machine learning, from the raw audio pipeline to the neural network that maps frequency patterns to chord names. They break down the real-time constraints, the trade-off between polyphony and accuracy, and how apps like Yousician and Simply Piano handle the challenge. Specific numbers include a typical model size of around 5 megabytes and inference under 50 milliseconds on a flagship phone. The episode also touches on how this compares to cloud-based approaches and why latency matters for practice. A concrete example: detecting a Cmaj7 in first inversion with a confidence threshold of 90%.

#PianoChordRecognition #OnDeviceAI #AudioProcessing #MachineLearning #RealTime #MusicTech #MobileAppDevelopment #iOS #Android #FourierTransform #NeuralNetworks #Yousician #SimplyPiano #Latency #Polyphony #Technology #FexingoBusiness #BusinessPodcast

Keep every episode free: buymeacoffee.com/fexingo

In episode 107 of Mobile Development with Fexingo, Lucas and Luna explore how on-device AI is enabling real-time guitar tablature transcription. They dive into the technical challenges of pitch detection, note segmentation, and latency constraints on mobile hardware. The hosts break down the specific machine learning models—like convolutional neural networks for spectrogram analysis and recurrent layers for temporal smoothing—that make this possible without cloud calls. They discuss a concrete use case: an app that transcribes a live guitar solo into tab notation in under 200 milliseconds, running entirely on an iPhone's Neural Engine or a Snapdragon-powered Android device. The conversation also covers training data curation, the trade-off between accuracy and power consumption, and why this is different from standard music recognition apps. A practical look at how mobile AI is democratizing music education and composition.

#OnDeviceAI #MobileApps #AppDevelopment #iOS #Android #GuitarTab #MusicTranscription #MachineLearning #RealTimeAI #NeuralNetworks #PitchDetection #MusicTech #EdgeComputing #DeepLearning #Technology #FexingoBusiness #BusinessPodcast #AI

Keep every episode free: buymeacoffee.com/fexingo

In this episode of Mobile Development with Fexingo, Lucas and Luna explore how on-device AI is enabling real-time emotion recognition in mobile apps. They dive into the technical challenges of running deep learning models on smartphones, the privacy advantages of keeping emotional data off the cloud, and a case study of a meditation app that uses facial expression analysis to adapt guided sessions. The hosts discuss the accuracy trade-offs, the ethical concerns around emotion AI, and how developers are training lightweight neural networks to detect subtle facial cues without draining battery. They also cover the latest advances in Apple's Core ML and Google's ML Kit for emotion detection, and why this technology is moving from experimental to production-ready. A must-listen for mobile developers and anyone curious about the intersection of AI and human emotion.

#EmotionRecognition #OnDeviceAI #MobileApps #CoreML #MLKit #FacialExpression #DeepLearning #PrivacyFirst #MeditationApp #AIethics #NeuralNetworks #RealTimeAI #iOSDev #AndroidDev #MachineLearning #Technology #FexingoBusiness #BusinessPodcast

Keep every episode free: buymeacoffee.com/fexingo

In this episode, Lucas and Luna explore how mobile apps are using on-device AI for real-time indoor navigation, a technology that helps users find their way inside large buildings like airports, hospitals, and shopping malls without relying on GPS. Lucas breaks down the core challenge: GPS signals are weak indoors, so apps must fuse data from Wi-Fi, Bluetooth beacons, and phone sensors to estimate precise location. He explains how machine learning models running on-device can combine signal fingerprints and step detection to achieve accuracy within a few meters. Luna asks about the practical hurdles, including the initial mapping effort and the risk of signal drift. Lucas shares how companies like Google and Apple are integrating these capabilities into their mapping SDKs, and how retail chains are using it for turn-by-turn directions to specific stores. The hosts discuss a real-world case: the Mall of America, which deployed indoor navigation across its 5.6 million square feet. They also touch on privacy advantages of on-device processing and future possibilities like AR overlays. The episode gives listeners a concrete understanding of how indoor navigation works under the hood and why it's becoming a standard feature in mobile apps.

#IndoorNavigation #OnDeviceAI #MobileApps #MachineLearning #WiFiFingerprinting #BluetoothBeacons #SensorFusion #GoogleMaps #AppleMaps #MallOfAmerica #Privacy #AR #Technology #FexingoBusiness #BusinessPodcast #iOS #Android #AppDevelopment

Keep every episode free: buymeacoffee.com/fexingo

In episode 104 of Mobile Development with Fexingo, Lucas and Luna explore how mobile apps are using on-device AI for real-time dance pose analysis. They break down the technology behind apps like DanceFit that can track a user's body movements and compare them to a professional dancer's, providing immediate feedback on timing, posture, and technique—all without sending video data to the cloud. The hosts unpack the key technical challenges: the switch from 2D keypoint detection to 3D pose estimation, the leap from 15 to 33 landmark points, and the clever use of temporal smoothing to handle rapid movements. They also discuss the delicate balance between accuracy and battery life on phones, and why the app's training data pool matters so much for performance across different body types and dance styles. A concrete look at how on-device AI is turning every phone into a personal dance coach.

#DancePoseAnalysis #OnDeviceAI #MobileApps #ComputerVision #PoseEstimation #DanceFit #MachineLearning #3DPose #KeypointDetection #RealTimeAI #FitnessTech #AIonMobile #MediaPipe #CoreML #MobileDevelopment #Technology #FexingoBusiness #BusinessPodcast

Keep every episode free: buymeacoffee.com/fexingo

In this episode of Mobile Development with Fexingo, hosts Lucas and Luna explore how on-device AI is enabling real-time guitar chord detection in mobile apps. They dive into the technical challenge of identifying chords from raw audio using small neural networks like convolutional neural networks (CNNs) and CRNNs, optimized to run on phones without a cloud round-trip. With demo apps like Chord AI and Yousician as examples, they discuss how audio preprocessing, chroma features, and model distillation make sub-100-millisecond chord recognition possible. Lucas explains the trade-offs between accuracy and latency, and why training on synthetic data from tools like PyTorch's audio libraries is key. The episode also touches on the business angle: how app developers monetize these features via freemium models and subscription tiers. A concrete, technical look at a niche but growing use case for edge AI in mobile music education.

#OnDeviceAI #MobileApps #GuitarChordDetection #MusicTech #iOSDev #AndroidDev #EdgeAI #MachineLearning #CNN #CRNN #AudioProcessing #ChromaFeatures #ModelDistillation #FexingoBusiness #BusinessPodcast #Technology #MobileDevelopment #RealTimeAI

Keep every episode free: buymeacoffee.com/fexingo

In this episode of Mobile Development with Fexingo, Lucas and Luna explore how on-device AI is transforming wine apps. They focus on the specific case of Vivino, which uses a neural network to identify wine labels from a photo and instantly returns tasting notes, ratings, and price comparisons. The hosts discuss the technical stack: a lightweight convolutional neural network trained on over 10 million label images, optimized for mobile CPUs and NPUs to run in under 200 milliseconds. They also cover challenges like varying lighting, bottle curvature, and label damage, and how data collection from users continuously improves accuracy. This concrete example shows how on-device AI delivers real-time utility without requiring a cloud round-trip, preserving user privacy and offline capability. A must-listen for developers building vision-based mobile features.

#OnDeviceAI #MobileDevelopment #WineRecognition #Vivino #ComputerVision #ConvolutionalNeuralNetwork #MachineLearning #iOS #Android #EdgeAI #NeuralProcessingUnit #LabelRecognition #RealTimeAI #PrivacyByDesign #OfflineCapability #Technology #FexingoBusiness #BusinessPodcast

Keep every episode free: buymeacoffee.com/fexingo

Lucas and Luna explore how on-device AI is enabling real-time sign language translation on smartphones. They dive into the technical challenges of capturing hand shapes, movement, and facial expressions simultaneously using the phone's camera and a lightweight neural network running entirely on the device. The episode focuses on a specific case: a 2025 research paper from Carnegie Mellon that achieved 93% accuracy on a 1,000-sign vocabulary without any cloud dependency. They discuss how models are compressed to fit under 100 MB, the role of Apple's Neural Engine and Google's Tensor Processing Unit, and why latency under 300 milliseconds is critical for natural conversation. The hosts also touch on current app examples like Google's Live Transcribe and Apple's SignTime, and where the technology still struggles — like regional dialect signs and two-handed gestures. A concrete, technical look at a rapidly maturing accessibility feature.

#OnDeviceAI #SignLanguage #AccessibilityTech #MobileApps #NeuralNetworks #ComputerVision #CarnegieMellon #AppleNeuralEngine #GoogleTensor #RealTimeTranslation #MachineLearning #iOS #Android #DeepLearning #GestureRecognition #Technology #FexingoBusiness #BusinessPodcast

Keep every episode free: buymeacoffee.com/fexingo

In this milestone 100th episode, Lucas and Luna dive into a controversial but real use of on-device AI: password cracking. They explore how apps like HashCat and John the Ripper leverage local neural networks to brute-force or guess passwords billions of times per second, all on a smartphone. The discussion centers on a 2025 study from Georgia Tech showing that a flagship Android phone can crack an 8-character alphanumeric password in under 4 hours using a distilled neural network. Lucas explains the technique—hash matching via a recurrent neural network trained on leaked password databases—and Luna questions the ethics and implications for user security. The hosts also touch on how this same technology is being repurposed for legitimate password recovery and security auditing. A sobering look at how powerful on-device AI has become.

#OnDeviceAI #PasswordCracking #MobileSecurity #Cybersecurity #NeuralNetworks #HashCat #JohnTheRipper #GeorgiaTech #Android #iOS #Technology #FexingoBusiness #BusinessPodcast #AI #PasswordRecovery #SecurityAudit #BruteForce #MachineLearning

Keep every episode free: buymeacoffee.com/fexingo