Jack Crossfire's blog View Details
Posted by Jack Crossfire | Today @ 12:27 AM | 208 Views
The embedded GPUs are out there. They're just 3x the price 2 years ago.

The lion kingdom believes the biggest improvements are going to come from improving the camera more than confusing power which no longer exists. Improving the exposure would improve every algorithm, so back light compensation is essential. Exposure could be adjusted so 99% of the pixels land above a certain minimum on the histogram. Unfortunately, the GeneralPlus has limited exposure control. Polling the video4linux2 driver, you get some settings.

id=980900 min=1 max=255 step=1 default=48 Brightness
id=980901 min=1 max=127 step=1 default=36 Contrast
id=980902 min=1 max=127 step=1 default=64 Saturation
id=980903 min=-128 max=127 step=1 default=0 Hue
id=980910 min=0 max=20 step=1 default=0 Gamma
id=98091b min=0 max=3 step=1 default=3 Sharpness
id=98091c min=0 max=127 step=1 default=8 Backlight Compensation

The only functional ones are contrast, saturation & backlight compensation. Backlight compensation is really some kind of brightness function. Saturation & contrast are some kind of software filters. The values have to be rewritten for every frame. Color definitely gives better face tracking than greyscale. The default backlight compensation is already as bright as possible. Changing saturation & backlight compensation from the default values gave no obvious improvement.


Face tracking could use higher resolution. There was also using a bigger target than a face by trying other demos in opencv-4.x/samples/dnn.

openpose.cpp ran at 1 frame every 30 seconds.

person-reid.cpp ran at 1.6fps. This requires prior detection of a person with openpose or YOLO. Then it tries to match it with a database.

object_detection.cpp ran at 1.8fps with the yolov4-tiny model from https://github.com/AlexeyAB/darknet#pre-trained-models This is a general purpose object detector.

Since only 1 in 4 frames are getting processed, there could be an alignment & averaging step. It would slow down the framerate. It wouldn't do well with fast motion.
Posted by Jack Crossfire | Jan 21, 2022 @ 03:12 PM | 2,541 Views
Face track run (1 hr 0 min 8 sec)


Complete face tracked run. It does a much better job than a lion manually pointing it. The only downside is not being able to manually point at something else.



Face track bloopers (2 min 49 sec)
...Continue Reading
Posted by Jack Crossfire | Jan 17, 2022 @ 09:24 PM | 12,812 Views
Face tracker 1 (8 min 14 sec)



The decision was made to transfer to using a face tracker instead of manually pointing the pole cam. Manual pointing would override the face tracker if the remote control was powered on, but it was expected to disappear. The lion kingdom found itself running more often without powering the servo, since carrying around the 2nd paw controller was a hassle. It was manely used for pointing the camera at the lion instead of viewing anything else. The most desirable shots were tracking the lion during a turn, but this wasn't possible with manual pointing.

The ideal tracker would run a pose tracker & face tracker. This can run reasonably well on a 4GB GPU. The face tracker would pick out the lion & the pose tracker would track any visible body part when the face wasn't visible. Unfortunately, the embedded GPUs of the past were all discontinued so the lion kingdom settled on just the opencv face tracker on a raspberry pi 4B.

Opencv face tracking is split into face detection & face recognition. Face recognition goes at 3fps while face detection goes at 7fps, so just the largest face is tracked. There is an optical flow step which keeps tracking the same face for a short time if a larger face appears. Optical flow works better at higher frame rates, but the best the pi can do is 7.8fps. Further optimization for slower modern confusers involved scaling down the video to 640x360.

The gopro 7 delays its HDMI feed...Continue Reading
Posted by Jack Crossfire | Jan 09, 2022 @ 11:15 PM | 30,776 Views
The old bumpers ended up delaminating & having CA glue release. The new bumpers use screws instead of glue & were printed at 250C instead of 220C to improve layer adhesion. The roll of filament is as hydrated as it gets. There's never been a filament drying method on the internet which costs less than a new roll of filament. The lion kingdom has dreamed of a solar heated method, but Calif*'s 3 months of tule fog are in full force.
Posted by Jack Crossfire | Jan 08, 2022 @ 11:08 PM | 34,303 Views
Many a lion has accidentally run 17 miles because target.com says it's only 7 miles away. Most impressive distance for a single battery. By not using the headlights or the speaker & not carrying any payload, it got under 200mAh/mile.

The lion kingdom managed to get the very last $14 wifi dongle in the entire country, but it's really more about the workout.
Posted by Jack Crossfire | Jan 04, 2022 @ 02:28 AM | 52,207 Views
Commercial products are normally a bare plastic enclosure surrounded by a silicone sock to provide waterproofing. Lions don't have enough room for silicone fabrication & the remote has some impossible protrusions.

Incrementally greater taping began, in an attempt to simulate a sock. There might be a way to seal in just the wettest part with a TPU clamshell sock. The TPU clamshell could be heat welded. The area around the grip tab is problematic. It's almost easier to make the LED side & the back side 2 different socks held on by a mild adhesive, tape, or some novel farstener. The sock could be a disposable thing with a weak seam in the middle or defined weld points suitable for cutting.
Posted by Jack Crossfire | Jan 01, 2022 @ 10:24 PM | 45,526 Views
So the last run of 2021 was an 11 miler in 42F. 2022 began with .25 miles before the paw controller died. The steering & throttle cross talked & it didn't charge anymore. Cracking it open revealed a pool of water under the PIC. So the water wicking idea didn't work. 11 miles in a sweaty lion paw must have caused a lot of condensation. Sitting in the charger didn't dry it off, probably because of the paper towel was soggy.

It did relatively better without the paper towel & with the fabric tape. It might be worth wrapping the fabric tape as thick as possible before the inductor dies. There could be a new charger with the electromagnet exposed instead of under .8mm of PLA.

Hot glue might be a better coating than polyurethane. It's been coming off easier for reworking & might just be 1 tool away from being as easy as air. It's not clear how well it keeps water out.
Posted by Jack Crossfire | Dec 30, 2021 @ 08:59 PM | 57,831 Views
So the battery stopped charging again. Opening it up revealed water condensation on the charging circuit & the inductor flopping around. Maybe it was worth conformal coating this board. This time, tried stuffing in a desiccant. Maybe it would wick water from the board, then dry out on the charger. The inductor needs rails on the side for it to slide into. The hot glue was coming off.

Reduced the charging current to try to reduce the inductor alignment, but it's unlikely the battery was ever getting over 300mA while the inductor is supposed to provide 500mA.
Posted by Jack Crossfire | Dec 25, 2021 @ 07:23 PM | 9,629 Views
With xmas rained out, it was finally time to disassemble the rudolf gear. It was better than any of the SUV decorations lions had seen for 20 years. The key was that string of 3V LEDs which was bought on a whim at Homeless Despot. No SUV has ever had lit antlers. It ended up quite easy to take down the antlers on the road. They never fell out on their own as feared. The solid core wire never broke, but they could have used a few wrappings of stranded wire.

Then there was the 15 mile run to Lowes to find a cheap red light bulb, the brief taste of rich millennial gootuber opulence before tearing it down, & the discovery that RGB bulbs don't produce their rated brightness from colored LEDs.

The only thing which might improve it is sleigh bells & even more red LEDs.
Posted by Jack Crossfire | Dec 21, 2021 @ 01:46 AM | 13,433 Views
Steering problems continued after the trace fix, the new linkages, & the paw controller fix.


Progressive failure of steering continued as the weather cooled. After a momentary loss of servo control, it regained its original heading, so the IMU was still working & the paw controller wasn't causing it to deviate. It always failed in higher elevations, never on the valley floor. That left a failure in the servo or mane board. The next idea was cleaning the mane board.

There's no reason for cleaning the mane board to make any difference because it wasn't affecting any other signals. The servo was rated as waterproof.


Another 30 miles in the cold damp didn't have any steering problems, so cleaning the board seemed to buy it some more time. Clenching PLA for 4 hours made the lion paws mighty raw.
Posted by Jack Crossfire | Dec 19, 2021 @ 10:01 PM | 13,336 Views
Roughly 1 year of 3D printed direct drive trucking since the concept took its 1st roll

https://www.rcgroups.com/forums/show...9955-1st-drive

That 1st drive was only .25 miles. 1 year later, ongoing steering excursions continue to be the mane problem.

The next problem is the inductive charging being piss poor. Part of it is water intrusion. That alone is still not as bad as the rusted JST connectors were.

The mane part is the alignment with wireless chargers having to be absolutely perfect. The original idea with inductive charging was to energize a very large surface on which a phone could be placed anywhere. Putting a phone anywhere on a desk was supposed to charge it. The reality is real chargers had such a small energized area, they relied on magnets to get the alignment perfect, which makes the gadgets heavy.

Thus, wireless charging is just about dead. Instead of becoming purely wireless, phones got a new USB connector. The wireless charging fiasco has made lions ponder a custom solution not designed by committee. The only obvious solution would be making a much larger charging coil, which would output a lot more power. It would still require a finicky protocol for connecting & disconnecting. Another idea is an array of smaller electromagnets, but it would be quite difficult without automated manufacturing. Finally, there's replacing the AAT3693 charging chip with something that took a wide variety of current or reducing the charging current.
Posted by Jack Crossfire | Dec 16, 2021 @ 05:49 PM | 8,916 Views
A bunch of low effort NFT's the lion kingdom found amusing contained a relevant story. In reality, new rover needs a lot more than sunshine, manely new tires, new servos, new wiring, new steering linkages, new batteries, new LEDs, cleaning, duct tape, replacement panels, water damage repair. Here the lion kingdom is with a JPG image which cost nothing to replicate, yet is valued at $6000 by some random authentic sounding word.


https://opensea.io/assets/0x495f9472...72713983131649
Posted by Jack Crossfire | Dec 11, 2021 @ 11:01 PM | 7,443 Views
Speed work on Thursday. Pay for it on Saturday.
Posted by Jack Crossfire | Dec 05, 2021 @ 08:27 PM | 55,448 Views
New ball links were printed at 220C. Careful use of the pliers got them on without delaminating. Ground a flat in the pushrod to aid the pliers. Tried lubricating the ball joints to try to keep any material from wearing off. Disappointing how fast PLA stretches & comes loose. Another idea might be TPU ball links.

Then came the well loved paw controller with water damage to the steering traces & a ground trace. Nearly the entire circuit is now potted. Removed an unused bodge wire. The power switch has a developing sticking in the off position which can be overcome by pulling it out. This too is probably water damage.
Posted by Jack Crossfire | Dec 05, 2021 @ 02:19 AM | 51,770 Views
As the weather cooled, the humidity hit 100%, & the lion sweated, full left glitches mounted. The lion kingdom believes it was water intrusion in the stick controller. Walking speed made the glitches go away. A steering linkage also came loose, as the PLA eventually stretched until it no longer gripped. This ended the trip after 16.9 miles.
Posted by Jack Crossfire | Dec 04, 2021 @ 03:49 AM | 37,360 Views
Running with Rudolf (2 min 41 sec)

Posted by Jack Crossfire | Nov 27, 2021 @ 02:07 AM | 27,442 Views
The left motor went to valhalla while slinging the DSLR 10 miles. Thus another day went to rewinding a motor with 20 turns of 26 AWG. The PETG motor mounts are taking the heat nicely, letting motor after motor fry. The retaining ring was documented to show future lion it was properly reinstalled. Those retaining ring pliers have proven essential.


Some notes during the replacement were crash damage to the bottom panel. The 1.6mm tread tire still delaminated. The next step would be 250C printing, but delamination doesn't seem to have any impact unless it picks up a branch.
Posted by Jack Crossfire | Nov 25, 2021 @ 01:29 AM | 16,782 Views
So the power burn with the DSLR + rudolf + headlights was 410mAh/mile. Tried to get the red glow reflecting from a tree tunnel. It shows up more when seen from behind. The EOS RP seems to have less noise in HDR mode. It takes 3 pictures with varying shutter speeds & might average areas which are properly exposed in all 3 areas.
Posted by Jack Crossfire | Nov 20, 2021 @ 07:12 PM | 16,560 Views
The amerikan people are in such war against pork, let alone meat, it's a rare day that rudolf brings back any pork flavor. Lions now manely buy grocery bags for garbage bags instead of putting food in. At what point are there so many restrictions on foods & grocery bags that it's no longer worth staying?
Posted by Jack Crossfire | Nov 17, 2021 @ 06:46 PM | 9,795 Views
The phone didn't do any better against TPU than it did against ziplock bags. The case got a piece of leather on the bottom, since that's where it slams down. There's still a lot of adhesive gunk near the bottom & scratching near the top because that's where it moves. The next meaningful improvement would require a PLA case with felt tape. There's no need for it to be flexible. The bottom needs a larger space to tape down. The bottom inside needs foam. The sides & especially the top need felt tape.

Otherwise, the goat wasn't fooled by the reindeer costume.

Goat vs Rudolf (2 min 4 sec)