In many commercial environments understanding the user's intention can lead to more engaging and intelligent user interactions. We looked at theme park photo kiosks where many people use their camera phones to capture their ride photos on preview displays. We believe that by identifying people with photo-taking intention and engaging them through intelligent UI can help reduce the instances of people opting for low quality but free screen capture. We built a prototype system called PhotoAct, using depth camera to recognize human postures and in real time infer people's photo-taking intentions. In this paper, we describe the system components, the detection algorithm, and present preliminary lab study results.