Scin

Scin

Scin is an app that utilizes a deep feed forward convolutional neural network to detect and diagnose both skin and plant diseases.

Modern Code, Artificial Intelligence

  • 0 Collaborators

  • 0 Followers

    Follow

Description

One out of every twenty people does not have access to medical facilities. Skin diseases and afflictions affect more than 80% of the world’s population and are sometimes a sign of internal problems. Skin diseases are primarily diagnosed visually, and then by more invasive procedures (dermoscopic analysis, biopsy, and histopathological examination). A similar issue arises with plants, with more than 30% of crops in less affluent areas dying due to diseases. Due to this loss of crops, one in nine people are suffering from chronic undernourishment.

Automatically detecting and diagnosing these lesions has been challenging, owing to the variable properties of each disease image. Deep convolutional neural networks (CNNs) are a new method of machine learning, one that is showing to be extremely promising at detecting images with real world variables. (lighting, focus, etc.)

In this project, I classified skin and plant ailments/diseases using a specially developed CNN, trained using only images of the conditions with only pixels and disease labels as inputs. I trained the CNN on a dataset of 200,000 clinical and horticultural images, consisting of 13 human diseases and 17 plant diseases. Outfitted on an IOS device, my application is capable of classifying skin and plant diseases will a level of competence comparable to dermatologists and plant pathologists. All the user must do is aim the camera of the smartphone towards the diseased area, and my application will provide a real-time diagnosis to the user by classifying the image using the CNN. The CNN achieves performance far above any other tested system, and its efficiency and ease of use will prove it to be a helpful tool for people around the world. There are currently 6,000,000,000 mobile subscriptions in place, so, therefore, my application could potentially provide low-cost universal access to vital diagnostics.

Gallery

Medium foto

Rafael S. created project Data Stream Mining

Medium c998e101 cead 4355 a005 b5614f7dfe88

Data Stream Mining

If the application data is very large and the recent data is the most important, by using data stream mining techniques we can extract knowledge to facilitate and simplify decision making. This project intends to study data streaming mining in internet networks, IoT sensors and social media. The machine learning algorithms need modifications to be adaptive and incremental. Modern code is also required to optimize computational resources.

Medium 11080561 764235763645452 7344272183143184984 o

Paulo P. created project CAMELBE3 - Machine learning and Intel RealSense for drowsiness detection

Medium 53c07649 3050 4e32 b839 23bee0a8a21b

CAMELBE3 - Machine learning and Intel RealSense for drowsiness detection

Machine learning and Intel RealSense technology for detecting drowsiness, drunkenness and even for predicting if the driver is going to change lanes.

The algorithm is running into an Upboard/laptop with Intel RealSense for capturing facial expressions and eye movements patterns. We tried to create our own version of the Nystagmus Test (not using fingers or pen, but windshield wipers). So far it's not that accurate, but we're working on it.

Medium 18342465 1561342793890595 4214787152430084956 n

Srivignessh P. created project Containerizing Deep Learning Workloads on Xeon Phi Cluster for AI Web Applications

Medium bae1ed0d 3a3a 40d6 8fba b5327e569eb3

Containerizing Deep Learning Workloads on Xeon Phi Cluster for AI Web Applications

Today's Web applications are data intensive and demand environments like tensorflow to execute the workloads. Hence, Containers are best suited to provide the framework and compute resources like CPU and memory for each workload. It decouples the app environment from the running machine/host and encapsulates all dependencies in a single portable unit. Nomad is a state of the art tool for scheduling Docker Containers. Test Model Workload is generated from Model Zoo for each framework.

See More

No users to show at the moment.

No users to show at the moment.