Dataset Information

Perceptual Learning of Object Recognition in Simulated Retinal Implant Perception - The Effect of Video Training.

ABSTRACT:

Purpose

Retinal implants (RIs) provide new vision for patients suffering from photoreceptor degeneration in the retina. The limited vision gained by RI, however, leaves room for improvement by training regimes.

Methods

Two groups of normal-sighted participants were respectively trained with videos or still images of daily objects in a labeling task. Object appearance was simulated to resemble RI perception. In Experiment 1, the training effect was measured as the change in performance during the training, and the same labeling task was conducted after 1 week to test the retention. In Experiment 2 with a different pool of participants, a reverse labeling task was included before (pre-test) and after the training (post-test) to show if the training effect could be generalized into a different task context.

Results

Both groups showed improved object recognition through training that was maintained for a week, and the video group showed better improvement (Experiment 1). Both groups showed improved object recognition in a different task that was maintained for a week, but the video group did not show better retention than the image group (Experiment 2).

Conclusions

Training with video materials leads to more improvement than training with still images in simulated RI perception, but this better improvement was specific to the trained task.

Translational relevance

We recommend videos as better training materials than still images for patients with RIs to improve object recognition when the task-goal is highly specific. We also propose here that achieving highly specific training goals runs the risk of limiting the generalization of the training effects.

SUBMITTER: Wang L

PROVIDER: S-EPMC8525839 | biostudies-literature | 2021 Oct

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Perceptual Learning of Object Recognition in Simulated Retinal Implant Perception - The Effect of Video Training.

Wang Lihui L Marek Nico N Steffen Johannes J Pollmann Stefan S

Translational vision science & technology 20211001 12

<h4>Purpose</h4>Retinal implants (RIs) provide new vision for patients suffering from photoreceptor degeneration in the retina. The limited vision gained by RI, however, leaves room for improvement by training regimes.<h4>Methods</h4>Two groups of normal-sighted participants were respectively trained with videos or still images of daily objects in a labeling task. Object appearance was simulated to resemble RI perception. In Experiment 1, the training effect was measured as the change in perform ...[more]

PMID: 34661623

Similar Datasets

Project description:Action video game (AVG) playing has been found to transfer to a variety of laboratory tasks in visual cognition. More recently, it has even been found to transfer to low-level visual "psychophysics tasks. This is unexpected since such low-level tasks have traditionally been found to be largely "immune" to transfer from another task, or even from the same task but a different stimulus attribute, e.g., motion direction. In this study, we set out to directly quantify transfer efficiency from AVG training to motion discrimination. Participants (n = 65) trained for 20 h on either a first-person active shooting video game, or a motion direction discrimination task with random dots. They were tested before, midway, and after training with the same motion task and an orientation discrimination task that had been shown to receive transfer from AVG training, but not from motion training. A subsequent control group (n = 18) was recruited to rule out any test-retest effect, by taking the same tests with the same time intervals, but without training. We found that improvement in motion discrimination performance was comparable between the AVG training and control groups, and less than the motion discrimination training group. We could not replicate the AVG transfer to orientation discrimination, but this was likely due to the fact that our participants were practically at chance for this task at all test points. Our study found no evidence, in either accuracy or reaction time, that AVG training transferred to motion discrimination. Overall, our results suggest that AVG training transferred little to lower-level visual skills, refining understanding of the mechanisms by which AVGs may affect vision. Protocol registration The accepted stage 1 protocol for this study can be found on the Open Science Framework at https://osf.io/zdv9c/?view_only=5b3b0c161dad448d9d1d8b14ce91ab11 . The stage 1 protocol for this Registered Report was accepted in principle on 01/12/22. The protocol, as accepted by the journal, can be found at: https://doi.org/10.17605/OSF.IO/ZDV9C.

Project description:Scene understanding and decomposition is a crucial challenge for intelligent systems, whether it is for object manipulation, navigation, or any other task. Although current machine and deep learning approaches for object detection and classification obtain high accuracy, they typically do not leverage interaction with the world and are limited to a set of objects seen during training. Humans on the other hand learn to recognize and classify different objects by actively engaging with them on first encounter. Moreover, recent theories in neuroscience suggest that cortical columns in the neocortex play an important role in this process, by building predictive models about objects in their reference frame. In this article, we present an enactive embodied agent that implements such a generative model for object interaction. For each object category, our system instantiates a deep neural network, called Cortical Column Network (CCN), that represents the object in its own reference frame by learning a generative model that predicts the expected transform in pixel space, given an action. The model parameters are optimized through the active inference paradigm, i.e., the minimization of variational free energy. When provided with a visual observation, an ensemble of CCNs each vote on their belief of observing that specific object category, yielding a potential object classification. In case the likelihood on the selected category is too low, the object is detected as an unknown category, and the agent has the ability to instantiate a novel CCN for this category. We validate our system in an simulated environment, where it needs to learn to discern multiple objects from the YCB dataset. We show that classification accuracy improves as an embodied agent can gather more evidence, and that it is able to learn about novel, previously unseen objects. Finally, we show that an agent driven through active inference can choose their actions to reach a preferred observation.

Dataset Information

Perceptual Learning of Object Recognition in Simulated Retinal Implant Perception - The Effect of Video Training.

Purpose

Methods

Results

Conclusions

Translational relevance

Publications

Perceptual Learning of Object Recognition in Simulated Retinal Implant Perception - The Effect of Video Training.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets