Rapid urbanization and global warming have increased heat stress in urban areas. This in turn makes using indoor space more compelling and leads to more energy consumption. Therefore, paying attention to outdoor spaces design with thermal comfort in mind becomes more important since outdoor spaces can host a variety of activities. This research aims to introduce a machine learning-based framework to predict the effects of different urban configurations (i.e. different greening configurations and types, different façade materials, and different urban geometry) on outdoor thermal comfort through training a pix2pix Convolutional generative adversarial network (cGAN) model. For the training of the machine learning model, a dataset consisting of 208 coupled pictures of input and output has been created. The simulation of this data has been carried out by ENVI-met. The resulting machine learning model had a Structural Similarity Index (SSIM) of 96% on the test dataset with the highest SSIM of 97.08 and lowest of 94.43 which shows the high accuracy of the model and it could have reached an answer in 3 s compared to the 30-min average time for ENVI-met simulation. The resulting model shows great promise for assisting researchers and urban designers in studying existing urban contexts or planning new developments. HIGHLIGHTS Machine learning use in outdoor thermal comfort assessment has been investigated. Vegetation, urban geometry, surface albedo, and water bodies have been studied parameters. Vegetation and street orientation have the highest and water bodies have the least impact on outdoor thermal comfort. Pix2pix algorithm implementation could create thermal comfort maps with 96% SSIM.

Purpose: In this study, a novel framework based on deep learning models is presented to assess energy and environmental performance of a given building space layout, facilitating the decision-making process at the early-stage design. Design/methodology/approach: A methodology using an image-based deep learning model called pix2pix is proposed to predict the overall daylight, energy and ventilation performance of a given residential building space layout. The proposed methodology is then evaluated by being applied to 300 sample apartment units in Tehran, Iran. Four pix2pix models were trained to predict illuminance, spatial daylight autonomy (sDA), primary energy intensity and ventilation maps. The simulation results were considered ground truth. Findings: The results showed an average structural similarity index measure (SSIM) of 0.86 and 0.81 for the predicted illuminance and sDA maps, respectively, and an average score of 88% for the predicted primary energy intensity and ventilation representative maps, each of which is outputted within three seconds. Originality/value: The proposed framework in this study helps upskilling the design professionals involved with the architecture, engineering and construction (AEC) industry through engaging artificial intelligence in human–computer interactions. The specific novelties of this research are: first, evaluating indoor environmental metrics (daylight and ventilation) alongside the energy performance of space layouts using pix2pix model, second, widening the assessment scope to a group of spaces forming an apartment layout at five different floors and third, incorporating the impact of building context on the intended objectives.