During Photovoltaic solar cells manufacturing,visual defects inspection is carried out to guarantee the product quality and life span of solar cells.Solar cell surface defects can be classified into cracks,cell breakages,finger interruptions etc.The presence of these defects decreases the efficiency and permeance of solar cells and sometimes may leave the solar cells fail to work.Therefore,timely detection of these defects is important to increase the durability and performance of solar cells.Industries often use manual defects inspection systems which are undoubtedly incompetent to detect defects efficiently and are quite expensive.However,due to the current advances in computer vision field,manual defect inspection can be substituted by automatic visual inspection methods.Although,the computer vision methods have made a great progress,the defects inspection in poly crystalline solar cells is still a challenging task because of the presence of random background patterns which are similar to defects.The unbalanced defect pixels distribution also makes the inspection critical.Furthermore,current methods are slow,inaccurate,inefficient and does not meet the industrial requirement.To detect the presence of cracks and finger interruption defects in solar cell images,a fast and highly efficient method is required to inspect cracks accurately.Thus,this thesis proposes a powerful yet effective method based on deep learning that can segment and detect subtle defects in solar cell electroluminescence images.The main contributions of this thesis are as follows,1)A novel end to end deep learning-based architecture is proposed for defects segmentation in poly crystalline solar cell electroluminescence images（EL）.In the proposed architecture we introduce a novel global attention to extract rich context information.Further,we modified the U-net by adding dilated convolution at both encoder and decoder side with skip connections from early layers to later layers at encoder side.Then the proposed global attention is incorporated into the modified U-net.The model is trained and tested on Photovoltaic electroluminescence 512x512 images dataset and the results are recorded using mean Intersection over union（IOU）.In experiments,we reported the results and made comparison between the proposed model and other state of the art methods.We demonstrate that the proposed method can segment various cracks robustly with smaller dataset and is computationally efficient.2)A new accurate defect detection method for photovoltaic electroluminescence（EL）images is proposed.The proposed algorithm leverages the advantage of multi attention network to efficiently extract the most important features and neglect the nonessential features during training.Firstly,we designed a channel attention to exploit contextual representations and spatial attention to effectively suppress background noise.Secondly,we incorporate both attention networks into modified U-net architecture and named it multi attention U-net（MAU-net）to extract effective multiscale features for defects inspection.Finally,we propose a hybrid loss which combines focal loss and dice loss aiming to solve two problems:a)overcome the class imbalance problem,and b)allowing the network to train with irregular image labels for some complex defects.The proposed multi attention U-net is evaluated on real photovoltaic EL images datasets using 5-fold cross validation technique.Experimental results demonstrate that the proposed network can segment and detect various complex defects correctly and can be adopted for real time industrial environment.