Application of active and transfer learning to wound image segmentation

Wound management is a fundamental task in standard clinical practice. Automated solutions already exist for humans, but there is a lack of applications on wound management for pets. The importance of a precise and efficient wound assessment is helpful to improve diagnosis and to increase the effectiveness of treatment plans for the chronic wounds. The goal of the research was to propose an automated pipeline capable of segmenting natural light-reflected wound images of animals. Two datasets composed by light-reflected images were used in this work: Deepskin dataset, 1564 human wound images obtained during routine dermatological exams, with 145 manual annotated images; Petwound dataset, a set of 290 wound photos of dogs and cats with 0 annotated images. Two implementations of U-Net Convolutioal Neural Network model were proposed for the automated segmentation. Active Semi-Supervised Learning techniques were applied for human-wound images to perform segmentation from 10% of annotated images. Then the same models were trained, via Transfer Learning, adopting an Active Semi- upervised Learning to unlabelled animal-wound images. The combination of the two training strategies proved their effectiveness in generating large amounts of annotated samples (94% of Deepskin, 80% of PetWound) with the minimal human intervention. The correctness of automated segmentation were evaluated by clinical experts at each round of training thus we can assert that the results obtained in this thesis stands as a reliable solution to perform a correct wound image segmentation. The use of Transfer Learning and Active Semi-Supervied Learning allows to minimize labelling effort from clinicians, even requiring no starting manual annotation at all. Moreover the performances of the model with limited number of parameters suggest the implementation of smartphone-based application to this topic, helping the future standardization of light-reflected images as acknowledge medical images.

Reinforcement learning over encrypted data

Reinforcement learning is a particular paradigm of machine learning that, recently, has proved times and times again to be a very effective and powerful approach. On the other hand, cryptography usually takes the opposite direction. While machine learning aims at analyzing data, cryptography aims at maintaining its privacy by hiding such data. However, the two techniques can be jointly used to create privacy preserving models, able to make inferences on the data without leaking sensitive information. Despite the numerous amount of studies performed on machine learning and cryptography, reinforcement learning in particular has never been applied to such cases before. Being able to successfully make use of reinforcement learning in an encrypted scenario would allow us to create an agent that efficiently controls a system without providing it with full knowledge of the environment it is operating in, leading the way to many possible use cases. Therefore, we have decided to apply the reinforcement learning paradigm to encrypted data. In this project we have applied one of the most well-known reinforcement learning algorithms, called Deep Q-Learning, to simple simulated environments and studied how the encryption affects the training performance of the agent, in order to see if it is still able to learn how to behave even when the input data is no longer readable by humans. The results of this work highlight that the agent is still able to learn with no issues whatsoever in small state spaces with non-secure encryptions, like AES in ECB mode. For fixed environments, it is also able to reach a suboptimal solution even in the presence of secure modes, like AES in CBC mode, showing a significant improvement with respect to a random agent; however, its ability to generalize in stochastic environments or big state spaces suffers greatly.

Continual learning on nano drones

Miniaturized flying robotic platforms, called nano-drones, have the potential to revolutionize the autonomous robots industry sector thanks to their very small form factor. The nano-drones’ limited payload only allows for a sub-100mW microcontroller unit for the on-board computations. Therefore, traditional computer vision and control algorithms are too computationally expensive to be executed on board these palm-sized robots, and we are forced to rely on artificial intelligence to trade off accuracy in favor of lightweight pipelines for autonomous tasks. However, relying on deep learning exposes us to the problem of generalization since the deployment scenario of a convolutional neural network (CNN) is often composed by different visual cues and different features from those learned during training, leading to poor inference performances. Our objective is to develop and deploy and adaptation algorithm, based on the concept of latent replays, that would allow us to fine-tune a CNN to work in new and diverse deployment scenarios. To do so we start from an existing model for visual human pose estimation, called PULPFrontnet, which is used to identify the pose of a human subject in space through its 4 output variables, and we present the design of our novel adaptation algorithm, which features automatic data gathering and labeling and on-device deployment. We therefore showcase the ability of our algorithm to adapt PULP-Frontnet to new deployment scenarios, improving the R2 scores of the four network outputs, with respect to an unknown environment, from approximately [−0.2, 0.4, 0.0,−0.7] to [0.25, 0.45, 0.2, 0.1]. Finally we demonstrate how it is possible to fine-tune our neural network in real time (i.e., under 76 seconds), using the target parallel ultra-low power GAP 8 System-on-Chip on board the nano-drone, and we show how all adaptation operations can take place using less than 2mWh of energy, a small fraction of the available battery power.