The goal of the CEFRL Workshop 2017 is to accelerate the study of compact and efficient feature representation and learning approaches in computer vision problems. We have entered the era of big data. The explosion of available visual data raises new challenges and opportunities. One major challenge is how to intelligently analyze and understand the unprecedented scale of visual data. Furthermore, mobile/wearable devices such as mobile phones and smart glasses are ubiquitous throughout our surroundings. Applications of feature representation technologies have to handle large-scale data or to run on mobile/wearable devices with limited computational capabilities and storage space, hence there is a growing need for feature descriptors that are fast to compute, memory efficient, and yet exhibiting good discriminability and robustness. This problem becomes more difficult when the data show various types of variations such as noise, illumination, scale, rotation and occlusion.