Optimal sensor height for multi-sensor capture

Hi there!

We are running a 5-sensor setup (4 forming a square + 1 higher in the middle of one of the sides) and we are wondering what the optimal height for the sensors would be.

In the tutorials I can see that the diagram shows the sensors at either chest height or head height. What would be better? We have recorded some tests and we are noticing poor quality of the captures in the lower half of the legs.

We also intend to record in which the actor will be seated the whole time, it would be better to lower the sensors right?

Hi, @FedericoGuardabrazo .

In our 5- and 6-sensor example configurations, the 4 “satellite” sensors are positioned roughly at waist-height, which can vary from subject to subject. On out R&D stage in Brooklyn which is usually set up in a semi-permanent general-purpose configuration for standing demos and captures, the sensors are 48" above the floor.

With a sitting subject, you may want to move the sensors down to seated waist level (~36") to get better coverage of the lower parts of the body as you have referenced, but ensure that at least one sensor can still see the top of the subject (crown of the head, tops of shoulders) so there are no gaps in the geometry data.

For highest quality, ensure that the sensors are as close to the subject as possible while still fitting the entire subject in each sensor’s color and depth frames - You can confirm this in Depthkit’s Record context and check each sensor one at a time to confirm what each sensor is capturing.

1 Like

You have to test what fits your setup and movement best. I have worked with 4 and 6, and 8 sensors up to 12 sensor setups … with 6 … I went for 4 around on the corners … then for face one, a bit closer and one over the stage … and for the 4 around, I had the two front left and right a bit higher looking more down and the 2 left and right in the back from lower angles to look under the arms …but I would say maybe even one central back sensor is OK … and then left right as cory says and then a face and a top sensor … you could try to pair to the face sensor with a better cinema style camera to get better colors … accordingly to your lighting quality … but as i said in the beginning somehow test and plan always testing days even direct before shooting … best even blocking the movement with the performer and then do test renders for fast workflow maybe first low quality mp4 fast renders to see the mesh quality and then if all is good then go for the highest quality export accordingly how you have planed it for your shooting.

Greetings Martn D.

1 Like