Video-lecture project weeks 1 and 2

The e-learning project for filming maths lectures is now underway (into its second week). You can see some of the initial results here and here. Read on for a summary of the early teething problems we have experienced and some of the solutions we have found.

For various reasons, weeks 1 and 2 of the project have been fraught with difficulties. In particular (see below) some of the recordings from week 1 were either fragmentary or unusable. We currently have usable versions of:

• Methods 3: the first hour of lecture 2 and all of lecture 3
• Complex analysis: the first hour of lecture 1, the first hour of lecture 2 and all of lecture 3.

Note that the even-numbered complex analysis lectures are interspersed with problem classes and we are only filming the first hour of these.

However, the biggest problem to date is that Lecturecast (the UCL video-lecture storage facility) doesn’t seem to like big files. This is why none of the lectures are currently up on Lecturecast. To get the feedback rolling in I have decided to upload some of the Methods 3 lectures to Youtube. For information, I got the go-ahead from UCL copyright to do this (under a CC-BY licence): the possible issue was that the UCL intellectual property policy states that UCL retains copyright where they have provided materials above-and-beyond the usual (in our case recording equipment) but I was told that this was not a problem so the performance rights are mine: thanks to June Hedges and Rod Digges for clarifying this.

The Methods 3 lectures 2 (part 1) and 3 are available here and here.

Some comments: The audio is poor, probably because I am having to speak loudly to a big lecture theatre (the mic is only recording, not broadcasting). The video starts a couple of seconds into the lecture (sometimes difficult to avoid). The second hour of lecture 2 was not captured because of lack of space on the 8GB SD card. We have now fixed this for future recordings.

Some other basic lessons we learnt:

1. Ensure that all equipment is fully charged before attempting to film a two hour lecture. The camera battery doesn't actually last a full two hours, but then nor does a two-hour lecture. Charging between the 9-10am and 11-12am lectures on Wednesdays only just recovers enough charge for the second lecture. We have now borrowed a cable which allows the camera to film whilst charging from e-learning and hopefully this will not be an issue again, but it led to some unusably fragmented videos early in week 1.

2. The Harrie Massey lecture theatre is badly suited for video capture. For a start (like the Chemistry lecture theatre) it is lacking in good power points to plug in the microphone base and we have had to run an extension cable from the front to the first row (posing something of a health and safety risk...) Moreover, there are bulky immovable objects directly in front of the whiteboards, rendering it difficult to film from the front row (where we need to be for electrical reasons). Bianca sensibly suggested filming slightly from one side to avoid these obstructions and this has improved matters.

3. An 8GB SD card is not enough for a two hour lecture. I have now bought two 16GB SD cards for this purpose. One must also remember to delete videos properly once they are downloaded: either by formatting the SD card or removing them from a command line – otherwise they go to a hidden trash file and take up space, which led to our not being able to film most of Methods 3 Lecture 1.

4. For uploading to Lecturecast, videos need to be encoded using h.264 as the video codec and aac as the audio codec. I have been compressing and encoding files using the free program FFMPEG (using version 1.2.2) but this is not the most user-friendly of programs and it is taking a while to get the videos small enough (but high enough quality) to upload to Lecturecast. Currently, the command
   $ ffmpeg -y -i ./raw/INPUT -b:v 500000-vcodec libx264 -vsync 1 -r 25 -acodec aac -strict -2 -ab 240k OUTPUT.mp4

   is producing decent output (an hour-long lecture taking up < 270MB). Sometimes the camera produces several output files, which can then be concatenated with
   $ ffmpeg -f concat -i list.txt -c copy OUTPUT.mp4

   where list.txt is a file containing a list in the format:
   file './OUTPUT1.mp4' file './OUTPUT2.mp4' etc.

5. Because the videos are not being uploaded to Lecturecast (and because I am keeping the unprocessed video-files in case I find I need an entirely different compression mechanism to get to Lecturecast) storage is becoming an issue. If I didn't have storage space at home, I would be panicking. My UCL desktop certainly doesn't have the storage capacity for even one of the unprocessed files and an 8GB memory stick won't store two. I also heartily recommend to anyone else running a similar project that you decide on an archiving/file-labelling system long in advance and stick to it religiously: the files from the camera are all called 0000n.MTS for some value of n and this can lead to confusion.