How to run NVIDIA NanoOWL tutorial - Revision history

Nico: /* Performance improvements */

2025-11-26T18:59:45Z

Performance improvements

← Older revision		Revision as of 18:59, 26 November 2025
Line 265:		Line 265:
	camera.set(cv2.CAP_PROP_BUFFERSIZE, 1)		camera.set(cv2.CAP_PROP_BUFFERSIZE, 1)
	logging.info("Loading predictor.")		logging.info("Loading predictor.")
	frame_count = 0
	predict_every_n_frames = 1 # predict every 15 frames

	def _read_and_encode_image():		def _read_and_encode_image():

2025-11-26T18:58:55Z

Performance improvements

← Older revision		Revision as of 18:58, 26 November 2025
Line 109:		Line 109:

	== Performance improvements ==		== Performance improvements ==

	<details>
	<summary>Click to show full code</summary>

	<syntaxhighlight lang="python">
	# ... your code here ...
	</syntaxhighlight>

	</details>


	* Observation: In the original implementation, both frame capture and model prediction were executed inside the same synchronous function ('''_read_and_encode_image'''). This caused the entire video pipeline to block while waiting for the model to finish inference, resulting in several seconds of latency in the browser.		* Observation: In the original implementation, both frame capture and model prediction were executed inside the same synchronous function ('''_read_and_encode_image'''). This caused the entire video pipeline to block while waiting for the model to finish inference, resulting in several seconds of latency in the browser.

2025-11-26T18:58:42Z

Performance improvements

← Older revision		Revision as of 18:58, 26 November 2025
Line 109:		Line 109:

	== Performance improvements ==		== Performance improvements ==

			<details>
			<summary>Click to show full code</summary>

			<syntaxhighlight lang="python">
			# ... your code here ...
			</syntaxhighlight>

			</details>


	* Observation: In the original implementation, both frame capture and model prediction were executed inside the same synchronous function ('''_read_and_encode_image'''). This caused the entire video pipeline to block while waiting for the model to finish inference, resulting in several seconds of latency in the browser.		* Observation: In the original implementation, both frame capture and model prediction were executed inside the same synchronous function ('''_read_and_encode_image'''). This caused the entire video pipeline to block while waiting for the model to finish inference, resulting in several seconds of latency in the browser.
Line 346:		Line 356:
	web.run_app(app, host=args.host, port=args.port)		web.run_app(app, host=args.host, port=args.port)
	</syntaxhighlight>		</syntaxhighlight>

	= Need Further Support? =		= Need Further Support? =

2025-11-26T18:57:19Z

How to run NVIDIA NanoOWL tutorial

← Older revision		Revision as of 18:57, 26 November 2025
Line 346:		Line 346:
	web.run_app(app, host=args.host, port=args.port)		web.run_app(app, host=args.host, port=args.port)
	</syntaxhighlight>		</syntaxhighlight>
			= Need Further Support? =

			'''📞 Book Consultation Call:''' [https://proventusnova.com/contact-us/ Show Calendar!]

			'''📩 Contact Via Email:''' [mailto:support@proventusnova.com support@proventusnova.com]

			'''🌐 Visit Our Website:''' [https://proventusnova.com ProventusNova.com]

2025-11-26T18:56:38Z

Adaptations

2025-11-26T18:39:39Z

Tutorial

← Older revision		Revision as of 18:39, 26 November 2025
Line 76:		Line 76:
	http://<board_ip>:7860		http://<board_ip>:7860
	</syntaxhighlight>		</syntaxhighlight>

			7. Now you are ready to test some prompts! Use the following as examples:
			* [a face]
			* [a face, a ball]
			* [a face [a nose, an eye]]

	== Adaptations ==		== Adaptations ==

2025-11-26T18:38:42Z

Tutorial

← Older revision		Revision as of 18:38, 26 November 2025
Line 67:		Line 67:

	You should see the following after the application starts:		You should see the following after the application starts:
			<syntaxhighlight lang="bash">
			======== Running on http://0.0.0.0:7860 ========
			(Press CTRL+C to quit)
			</syntaxhighlight>

			6. On your host PC, open browser to:
			<syntaxhighlight lang="bash">
			http://<board_ip>:7860
			</syntaxhighlight>

	== Adaptations ==		== Adaptations ==

2025-11-26T18:34:02Z

Set up

← Older revision		Revision as of 18:34, 26 November 2025
Line 25:		Line 25:
	ssh <user>@<board_ip>		ssh <user>@<board_ip>
	</syntaxhighlight>		</syntaxhighlight>

			Now you should be inside the board.

	== Tutorial ==		== Tutorial ==

2025-11-26T18:32:25Z

Set up

2025-11-26T18:30:52Z

Tutorial

← Older revision		Revision as of 18:30, 26 November 2025
Line 26:		Line 26:

	== Tutorial ==		== Tutorial ==

	This tutorial was tested on an NVIDIA Jetson Orin NX (16 GB)
	1. Clone and setup '''jetson-containers'''		1. Clone and setup '''jetson-containers'''