In the last decade, Autism Spectrum Disorder (ASD) prevalence rate has significantly increased, which consequently led to the expansion of research and expenditure in the field, predominantly focusing on searching for the cause. In a typical classroom scenario, working with children with ASD very often requires 1:1 teacher to child ratio, which makes it very expensive and difficult to implement. Serious games have been utilised as a medium for teaching various developmental skills, such as social interaction, speech, motor skills development, emotion recognition, and other basic concepts. Designing serious games for ASD population differs from other games and even other serious games significantly. It requires a holistic approach with extensive knowledge and expertise from fields other than computer science, such as psychology, sociology and cognitive science. However, once harnessed correctly, such games can be used by children with ASD on their own time, with or without supervision and they can be educational. In addition, they can adjust the appropriate pace while at the same time providing feedback in form of reinforcement and correction. Applying the rules of science of learning and teaching, one can design games that are educational for all types of learners, including children with ASD. In this paper, two independent user studies have been conducted, demonstrating how serious gaming and e-learning principles can be harnessed in order to intervene, develop or strengthen pivotal developmental skills, like learning novel vocabulary, counting, identifying numbers and colours, and responding to inference questions. We have tested the educational e-book with children diagnosed with ASD and with typically developing children to assess skill acquisition in native language for children with ASD and in English, a foreign language, for typically developing children to demonstrate the educational aspect of the game for all types of learners. We showed that the same e-book in two languages can be used for teaching different types of learners through a fun and engaging medium.

Image compression standards rely on predictive coding, transform coding, quantization and entropy coding, in order to achieve high compression performance. Very recently, deep generative models have been used to optimize or replace some of these operations, with very promising results. However, so far no systematic and independent study of the coding performance of these algorithms has been carried out. In this paper, for the first time, we conduct a subjective evaluation of two recent deep-learning-based image compression algorithms, comparing them to JPEG 2000 and to the recent BPG image codec based on HEVC Intra. We found that compression approaches based on deep auto-encoders can achieve coding performance higher than JPEG 2000, and sometimes as good as BPG. We also show experimentally that the PSNR metric is to be avoided when evaluating the visual quality of deep-learning-based methods, as their artifacts have different characteristics from those of DCT or wavelet-based codecs. In particular, images compressed at low bitrate appear more natural than JPEG 2000 coded pictures, according to a no-reference naturalness measure. Our study indicates that deep generative models are likely to bring huge innovation into the video coding arena in the coming years.

The demand for very high-resolution video content in entertainment services (4K, 8K, panoramic, 360 VR) puts an increasing load on the distribution network. In order to reduce the network usage in existing delivery infrastructure for such services while keeping a good quality of experience, dynamic spatial video adaptation at the client side is seen as a key feature, and is actively investigated by academics and industrials. However, the impact of spatial adaptation on quality perception is not clear. In this paper, we propose a methodology for the evaluation of such adapted content, conduct a series of perceived quality measurements and discuss results showing potential benefits and drawbacks of the technique. Based on our results, we also propose a signaling mechanism in MPEG-DASH to assist the client in its spatial adaptation logic.