A Year's Walk Through the Neighborhood

A major project for 2012 was taking the photographs for a one-year zoomwalk through my neighborhood. This video is constructed from 1,768 photos taken between January 1st and December 31st, 2012. The date each photo was taken is in the lower right-hand corner of the frame.

I took two steps between photos, and shot between 6 and 10 photos per day. Also, I was away for several one to two week periods during the year. As you might suspect, despite my efforts, taking photographs in small chunks on separate days or even weeks increases the "wiggle" of the video. Going around curves was tricky, and I learned to take only one step per photo in such situations.

Note that the year started with three Christmas Pigs in the yard, but ended with four.

If you would like to see the video in its full resolution (1280 x 720), click here (you will be taken to the Vimeo web site).



Technique Tweaks
I mostly recently discussed the frame alignment and morphing techniques used in this earlier post. Since then, there's been one major tweak to the frame alignment algorithm.

Last year, for a sequence of photos 1-2-3-4-5-6, photo #2 would be an anchor to which #1 and #3 would be compared, #5 would be an anchor to which #4 and #6 would be compared, and so forth. The choice of anchors was fixed and immutable. Now, the anchors are chosen based on prior results. For example, let's say that the comparison of #3 to the anchor, #2,  succeeded; then the next anchor would still be #5 (to compare #4 and #6 against). But if the alignment of #3 and #2 failed -- the change to #3 was too great -- then the next anchor is #4, not #5. Photos #3 and #5 would be compared to #4. Thus there is a second chance to align #3 to an adjacent photo if the first attempt fails. With this improvement, fully half of the photos in this zoomwalk were aligned.

And here is a photo to feed a thumbnail to the Recent Posts and Popular Posts widgets.

Snow Flakes, Shovels, and Plows: A Timelapse

I recently bought a Bushnell TrophyCam, and its first extended test was looking out from our front porch when we expected a significant snowfall. We ended up with 5 inches.

The camera was set to record a picture every 5 minutes in addition to a picture when triggered by its motion sensor. It switches between black-and-white images and color depending on the light level. I think it turned out pretty well!

Here's a still photo to satisfy blogger thumbnailing,

and here's the video.


Snow Flakes, Shovels, and Plows: Timelapse from Ben Branch on Vimeo.

PlantCam Timelapse with Clivia and Snowfall

In addition to a couple of books, at Christmas time I received a PlantCam, a time-lapse camera from the same outfit that makes the motion-sensitive BirdCam that we've used for the last couple of years. You've seen some pictures from that unit, including this post. With this blog entry I will subject you to my first time-lapse "compositions." No doubt there will be a learning curve.

Note: the YouTube clips default to a 360p setting, at least for me, but the original material is 480 pixels high. The clips will look much crisper if you go to the 360p that shows up on the right side of the YouTube playback controls and click to change it to 480p. Unfortunately, the clickable 360p shows up only after playback begins! If I ever find a way around this ugliness, I will update this post.

My first subject was a large, nay, huge clivia that grows near our first-floor bay window. In late December Clivia (that's her proper name) starting pushing up a stalk, and I set the PlantCam to take a frame once every 5 minutes, with this clip recording from December 31st to January 17th.

As you can see, the camera position with regard to the sun was poor, causing washout and flaring. (I threw out the worst of the glare-burned frames.)  But it did reveal that Clivia does a lot of growing at night, of which there is a surplus at year's end, with the first frame of each new day showing a jump in the stalk.

The next clip is of the snowfall -- well, the first part of it -- of January 20th. The camera was positioned on a tripod in a protected corner of the front porch, and set to grab one frame per minute. After a while the snow starts to accumulate on the street, and then the snow on the bushes in the immediate foreground piles up. You will notice that I changed the PlantCam setting so that it no longer smacked an intrusive imprint  across the bottom of the frame; instead I wrote a script using ImageMagick, an open-source tool, to create a watermark in the lower left corner with the date and time.

Technical note: the date and time were extracted from each frame of the clip, specifically, from the EXIF data that every modern digital camera includes in a JPEG file. I don't use the PlantCam's ability to create the movie within the camera; first, I would rather have the opportunity to process each frame, as with the watermark, and second, the PlantCam's automatic movie generation cannot handle a large number of frames. I again used an open-source tool, ffmpeg, to generate a movie from the many still images.

The final and best clip is of the very sunny day after the snowfall. The light and shadows make this clip, recorded at one frame every 30 seconds, great fun to watch. Yours truly is one of the early shovelers.

I'll continue to experiment with the PlantCam, and come spring and summer, there will be plenty of daylight.