One suggestion that has been made for handling different frame rates in the source and rendition is sampling timestamps instead of frame numbers comparing coincident frames in the source and rendition (occurring at the same timestamp). This would require adding a quantization step to grab coincident frames in the source and rendition. The downside of this approach is that when either upsampling or downsampling s.t. the rendition FPS is not a divisor of the source FPS, not all of the rendition frames will be eligible for sampling. However, given these two cases are not very common, it could be useful to experiment with this approach to confirm that it at least addresses the common case of downsampling where the rendition FPS is a divisor of the source FPS (i.e. 60fps -> 30fps).
One suggestion that has been made for handling different frame rates in the source and rendition is sampling timestamps instead of frame numbers comparing coincident frames in the source and rendition (occurring at the same timestamp). This would require adding a quantization step to grab coincident frames in the source and rendition. The downside of this approach is that when either upsampling or downsampling s.t. the rendition FPS is not a divisor of the source FPS, not all of the rendition frames will be eligible for sampling. However, given these two cases are not very common, it could be useful to experiment with this approach to confirm that it at least addresses the common case of downsampling where the rendition FPS is a divisor of the source FPS (i.e. 60fps -> 30fps).