Face detection constitutes a key visual information analysis task in MachineLearning. The rise of Big Data has resulted in the accumulation of a massivevolume of visual data which requires proper and fast analysis. Deep Learn-ing methods are powerful approaches towards this task as training with largeamounts of data exhibiting high variability has been shown to significantly en-hance their effectiveness, but often require expensive computations and lead tomodels of high complexity. When the objective is to analyse visual content inmassive datasets the complexity of the model becomes crucial to the success ofthe model. In this paper, a light-weight deep Convolutional Neural Network(CNN) is introduced for the purpose of face detection, designed with a view tominimize training and testing time, and outperforms previously published deepconvolutional networks in this task in terms of both effectiveness and efficiency.The model consists of 76.375 free parameters whereas most of the competitiveones consisted of millions of parameters. To train this lightweight deep networkwithout compromising its efficiency, a new training method of progressive pos-itive and hard negative sample mining is introduced and shown to drasticallyimprove training speed and accuracy. Additionally, a separate deep networkwas trained to detect individual facial features and a model that combines theoutputs of the two networks was created and evaluated. Both methods are ableto detect faces under severe occlusion and unconstrained pose variation andmeet the difficulties and the large variations of large scale real-world, real-timeface detection.