{"id":39360,"date":"2026-09-12T16:41:58","date_gmt":"2026-09-12T08:41:58","guid":{"rendered":"http:\/\/hazlina-backup-39360"},"modified":"2026-09-12T16:41:58","modified_gmt":"2026-09-12T08:41:58","slug":"umpsa-stem-lab-esp32-microcontroller-programming-2026","status":"publish","type":"post","link":"\/hazlina-blog\/2026\/09\/12\/umpsa-stem-lab-esp32-microcontroller-programming-2026\/","title":{"rendered":"UMPSA STEM Lab \u2013 Edge Intelligence (Image Processing) on ESP32 2026"},"content":{"rendered":"<p><span class=\"rnc2Gd\" data-ast-node-id=\"1\">Deploying artificial intelligence directly on compact microcontrollers\u2014often termed\u00a0<\/span><strong><span class=\"rnc2Gd\" data-ast-node-id=\"2\">Edge Intelligence<\/span><\/strong><span class=\"rnc2Gd\" data-ast-node-id=\"3\">\u2014represents a major leap in embedded engineering<\/span><span class=\"rnc2Gd\" data-ast-node-id=\"4\">.\u00a0<\/span><span class=\"rnc2Gd\" data-ast-node-id=\"5\">This <strong>U<\/strong><\/span><strong><span class=\"rnc2Gd\" data-ast-node-id=\"6\">MPSA STEM Lab<\/span><\/strong><span class=\"rnc2Gd\" data-ast-node-id=\"7\">\u00a0module provides a hands-on roadmap for transitioning from basic software-based computer vision to standalone, hardware-integrated AI models capable of performing real-time image processing, feature extraction, and object classification on low-power ESP microcontroller hardware<\/span><span class=\"rnc2Gd\" data-ast-node-id=\"8\">.<\/span><\/p>\n<p><strong><span class=\"rnc2Gd\" data-ast-node-id=\"9\">The Shift to Edge Intelligence<\/span><\/strong><\/p>\n<p><span class=\"rnc2Gd\" data-ast-node-id=\"10\">Traditional computer vision solutions rely on cloud servers or heavy laptop GPUs to process video streams. Edge Intelligence moves processing power directly to the physical sensor level<\/span><span class=\"rnc2Gd\" data-ast-node-id=\"11\">. <\/span><span class=\"rnc2Gd\" data-ast-node-id=\"12\">In this module, participants move beyond pre-packaged datasets to create, train, and deploy bespoke computer vision pipelines designed specifically for embedded micro-processors<\/span><span class=\"rnc2Gd\" data-ast-node-id=\"13\">.<\/span><\/p>\n<div class=\"rnc2Gd Wv1ane i03hic\" data-open=\"false\" data-type=\"13\" data-url=\"https:\/\/drive.google.com\/file\/d\/191HA1cB9vQ-_zSObj55yHeCq781M6VrH\/view?usp=drive_web\" data-edit-type=\"2\" data-anchor-id=\"ucc-174\" data-anchor-corner=\"5\" data-link-preview-card-corner=\"6\" data-show-thumbnail=\"true\" data-show-preview-delay-ms=\"0\" data-hide-preview-delay-ms=\"0\" data-is-citation-source=\"false\" data-is-internal-link=\"false\" data-show-manually=\"false\" data-animate-entry=\"false\" data-prevent-default-link-navigation=\"false\" data-link-preview-card-classes=\"MmSINd\"><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-large wp-image-39363\" src=\"\/hazlina\/wp-content\/uploads\/2026\/09\/Screenshot-2026-09-12-at-4.57.34\u202fPM-1024x292.png\" alt=\"\" width=\"840\" height=\"240\" srcset=\"\/hazlina\/wp-content\/uploads\/2026\/09\/Screenshot-2026-09-12-at-4.57.34\u202fPM-1024x292.png 1024w, \/hazlina\/wp-content\/uploads\/2026\/09\/Screenshot-2026-09-12-at-4.57.34\u202fPM-300x85.png 300w, \/hazlina\/wp-content\/uploads\/2026\/09\/Screenshot-2026-09-12-at-4.57.34\u202fPM-768x219.png 768w, \/hazlina\/wp-content\/uploads\/2026\/09\/Screenshot-2026-09-12-at-4.57.34\u202fPM-1200x342.png 1200w, \/hazlina\/wp-content\/uploads\/2026\/09\/Screenshot-2026-09-12-at-4.57.34\u202fPM.png 1320w\" sizes=\"auto, (max-width: 709px) 85vw, (max-width: 909px) 67vw, (max-width: 1362px) 62vw, 840px\" \/><\/div>\n<p><strong><span class=\"rnc2Gd\" data-ast-node-id=\"15\">Image Processing &amp; Data Engineering Discipline<\/span><\/strong><\/p>\n<p><span class=\"rnc2Gd\" data-ast-node-id=\"16\">A fundamental lesson of embedded machine learning is that an AI model is only as effective as the data used to train i<\/span><span class=\"rnc2Gd\" data-ast-node-id=\"17\">t.<\/span><\/p>\n<div class=\"rnc2Gd Wv1ane i03hic\" data-open=\"false\" data-type=\"13\" data-url=\"https:\/\/drive.google.com\/file\/d\/1kdLwzMaJyrq0JC1zyDm8wsOVOvWLAl6O\/view?usp=drive_web\" data-edit-type=\"2\" data-anchor-id=\"ucc-178\" data-anchor-corner=\"5\" data-link-preview-card-corner=\"6\" data-show-thumbnail=\"true\" data-show-preview-delay-ms=\"0\" data-hide-preview-delay-ms=\"0\" data-is-citation-source=\"false\" data-is-internal-link=\"false\" data-show-manually=\"false\" data-animate-entry=\"false\" data-prevent-default-link-navigation=\"false\" data-link-preview-card-classes=\"MmSINd\"><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-large wp-image-39365\" src=\"\/hazlina\/wp-content\/uploads\/2026\/09\/Screenshot-2026-09-12-at-4.55.10\u202fPM-1024x573.png\" alt=\"\" width=\"840\" height=\"470\" srcset=\"\/hazlina\/wp-content\/uploads\/2026\/09\/Screenshot-2026-09-12-at-4.55.10\u202fPM-1024x573.png 1024w, \/hazlina\/wp-content\/uploads\/2026\/09\/Screenshot-2026-09-12-at-4.55.10\u202fPM-300x168.png 300w, \/hazlina\/wp-content\/uploads\/2026\/09\/Screenshot-2026-09-12-at-4.55.10\u202fPM-768x429.png 768w, \/hazlina\/wp-content\/uploads\/2026\/09\/Screenshot-2026-09-12-at-4.55.10\u202fPM-1200x671.png 1200w, \/hazlina\/wp-content\/uploads\/2026\/09\/Screenshot-2026-09-12-at-4.55.10\u202fPM.png 1756w\" sizes=\"auto, (max-width: 709px) 85vw, (max-width: 909px) 67vw, (max-width: 1362px) 62vw, 840px\" \/><\/div>\n<div data-open=\"false\" data-type=\"13\" data-url=\"https:\/\/drive.google.com\/file\/d\/1kdLwzMaJyrq0JC1zyDm8wsOVOvWLAl6O\/view?usp=drive_web\" data-edit-type=\"2\" data-anchor-id=\"ucc-178\" data-anchor-corner=\"5\" data-link-preview-card-corner=\"6\" data-show-thumbnail=\"true\" data-show-preview-delay-ms=\"0\" data-hide-preview-delay-ms=\"0\" data-is-citation-source=\"false\" data-is-internal-link=\"false\" data-show-manually=\"false\" data-animate-entry=\"false\" data-prevent-default-link-navigation=\"false\" data-link-preview-card-classes=\"MmSINd\"><\/div>\n<div data-open=\"false\" data-type=\"13\" data-url=\"https:\/\/drive.google.com\/file\/d\/1kdLwzMaJyrq0JC1zyDm8wsOVOvWLAl6O\/view?usp=drive_web\" data-edit-type=\"2\" data-anchor-id=\"ucc-178\" data-anchor-corner=\"5\" data-link-preview-card-corner=\"6\" data-show-thumbnail=\"true\" data-show-preview-delay-ms=\"0\" data-hide-preview-delay-ms=\"0\" data-is-citation-source=\"false\" data-is-internal-link=\"false\" data-show-manually=\"false\" data-animate-entry=\"false\" data-prevent-default-link-navigation=\"false\" data-link-preview-card-classes=\"MmSINd\"><\/div>\n<div data-open=\"false\" data-type=\"13\" data-url=\"https:\/\/drive.google.com\/file\/d\/1kdLwzMaJyrq0JC1zyDm8wsOVOvWLAl6O\/view?usp=drive_web\" data-edit-type=\"2\" data-anchor-id=\"ucc-178\" data-anchor-corner=\"5\" data-link-preview-card-corner=\"6\" data-show-thumbnail=\"true\" data-show-preview-delay-ms=\"0\" data-hide-preview-delay-ms=\"0\" data-is-citation-source=\"false\" data-is-internal-link=\"false\" data-show-manually=\"false\" data-animate-entry=\"false\" data-prevent-default-link-navigation=\"false\" data-link-preview-card-classes=\"MmSINd\"><img loading=\"lazy\" decoding=\"async\" class=\"alignnone wp-image-39366 size-full\" src=\"\/hazlina\/wp-content\/uploads\/2026\/09\/watermarked_img_7495501931007671599-e1789203766507.jpg\" alt=\"\" width=\"770\" height=\"683\" srcset=\"\/hazlina\/wp-content\/uploads\/2026\/09\/watermarked_img_7495501931007671599-e1789203766507.jpg 770w, \/hazlina\/wp-content\/uploads\/2026\/09\/watermarked_img_7495501931007671599-e1789203766507-300x266.jpg 300w, \/hazlina\/wp-content\/uploads\/2026\/09\/watermarked_img_7495501931007671599-e1789203766507-768x681.jpg 768w\" sizes=\"auto, (max-width: 709px) 85vw, (max-width: 909px) 67vw, (max-width: 984px) 61vw, (max-width: 1362px) 45vw, 600px\" \/><\/div>\n<div class=\"gWqWw rnc2Gd\"><\/div>\n<p><i><span class=\"rnc2Gd\" data-ast-node-id=\"18\">Figure 1: The compact ESP32 camera board featuring an integrated 2 megapixel camera sensor and Wi-Fi capabilit<\/span><span class=\"rnc2Gd\" data-ast-node-id=\"19\">y.<\/span><\/i><\/p>\n<h3><span class=\"rnc2Gd\" data-ast-node-id=\"20\">The Thumbnail Test &amp; Image Resolution<\/span><\/h3>\n<ol>\n<li style=\"list-style-type: none\">\n<ol>\n<li style=\"list-style-type: none\">\n<ol class=\"gdAMzb\">\n<li class=\"rnc2Gd\"><strong><span class=\"rnc2Gd\" data-ast-node-id=\"21\">Pixel Reduction<\/span><\/strong><span class=\"rnc2Gd\" data-ast-node-id=\"22\">\u00a0Raw images captured by the sensor are downscaled to 96 x 96 pixel grayscale or RGB array<\/span><span class=\"rnc2Gd\" data-ast-node-id=\"23\">s.<\/span><\/li>\n<li class=\"rnc2Gd\"><strong><span class=\"rnc2Gd\" data-ast-node-id=\"24\">Visual Clarity<\/span><\/strong><span class=\"rnc2Gd\" data-ast-node-id=\"25\">\u00a0Before training, images are evaluated via the &#8220;thumbnail test&#8221; to ensure visual features remain recognizable when shrunk to stamp-sized dimension<\/span><span class=\"rnc2Gd\" data-ast-node-id=\"26\">s.<\/span>\n<div class=\"rnc2Gd Wv1ane i03hic\" data-open=\"false\" data-type=\"13\" data-url=\"https:\/\/drive.google.com\/file\/d\/1kdLwzMaJyrq0JC1zyDm8wsOVOvWLAl6O\/view?usp=drive_web\" data-edit-type=\"2\" data-anchor-id=\"ucc-181\" data-anchor-corner=\"5\" data-link-preview-card-corner=\"6\" data-show-thumbnail=\"true\" data-show-preview-delay-ms=\"0\" data-hide-preview-delay-ms=\"0\" data-is-citation-source=\"false\" data-is-internal-link=\"false\" data-show-manually=\"false\" data-animate-entry=\"false\" data-prevent-default-link-navigation=\"false\" data-link-preview-card-classes=\"MmSINd\"><\/div>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-large wp-image-39364\" src=\"\/hazlina\/wp-content\/uploads\/2026\/09\/Screenshot-2026-09-12-at-4.55.45\u202fPM-1024x572.png\" alt=\"\" width=\"840\" height=\"469\" srcset=\"\/hazlina\/wp-content\/uploads\/2026\/09\/Screenshot-2026-09-12-at-4.55.45\u202fPM-1024x572.png 1024w, \/hazlina\/wp-content\/uploads\/2026\/09\/Screenshot-2026-09-12-at-4.55.45\u202fPM-300x168.png 300w, \/hazlina\/wp-content\/uploads\/2026\/09\/Screenshot-2026-09-12-at-4.55.45\u202fPM-768x429.png 768w, \/hazlina\/wp-content\/uploads\/2026\/09\/Screenshot-2026-09-12-at-4.55.45\u202fPM-1200x670.png 1200w, \/hazlina\/wp-content\/uploads\/2026\/09\/Screenshot-2026-09-12-at-4.55.45\u202fPM.png 1758w\" sizes=\"auto, (max-width: 709px) 85vw, (max-width: 909px) 67vw, (max-width: 1362px) 62vw, 840px\" \/><\/li>\n<\/ol>\n<\/li>\n<\/ol>\n<\/li>\n<\/ol>\n<h3><span class=\"rnc2Gd\" data-ast-node-id=\"27\">Essential Rules for Dataset Quality<\/span><\/h3>\n<ol>\n<li style=\"list-style-type: none\">\n<ol>\n<li style=\"list-style-type: none\">\n<ol>\n<li style=\"list-style-type: none\">\n<ol class=\"gdAMzb\">\n<li class=\"rnc2Gd\"><strong><span class=\"rnc2Gd\" data-ast-node-id=\"28\">Single Variable Changes<\/span><\/strong><span class=\"rnc2Gd\" data-ast-node-id=\"29\">\u00a0Change only one condition (angle, lighting, position) between consecutive photo captures to maximize information densit<\/span><span class=\"rnc2Gd\" data-ast-node-id=\"30\">y.<\/span><\/li>\n<li class=\"rnc2Gd\"><strong><span class=\"rnc2Gd\" data-ast-node-id=\"31\">Background Separation<\/span><\/strong><span class=\"rnc2Gd\" data-ast-node-id=\"32\">\u00a0Avoid shooting all target objects on identical surfaces. Otherwise, the model learns background cues rather than object feature<\/span><span class=\"rnc2Gd\" data-ast-node-id=\"33\">s.<\/span><\/li>\n<li class=\"rnc2Gd\"><strong><span class=\"rnc2Gd\" data-ast-node-id=\"34\">Consistent Label Formatting<\/span><\/strong><span class=\"rnc2Gd\" data-ast-node-id=\"35\">\u00a0Labels must follow strict naming rules (lower case, no spaces, e.g.,\u00a0<\/span><code class=\"dfqOJc\">red_chili<\/code><span class=\"rnc2Gd\" data-ast-node-id=\"37\">) to prevent duplicate target classe<\/span><span class=\"rnc2Gd\" data-ast-node-id=\"38\">s.<\/span>\n<div class=\"rnc2Gd Wv1ane i03hic\" data-open=\"false\" data-type=\"13\" data-url=\"https:\/\/drive.google.com\/file\/d\/1sFTC8DDfPVe-cv68qeAIohGcg3NqKtW6\/view?usp=drive_web\" data-edit-type=\"2\" data-anchor-id=\"ucc-184\" data-anchor-corner=\"5\" data-link-preview-card-corner=\"6\" data-show-thumbnail=\"true\" data-show-preview-delay-ms=\"0\" data-hide-preview-delay-ms=\"0\" data-is-citation-source=\"false\" data-is-internal-link=\"false\" data-show-manually=\"false\" data-animate-entry=\"false\" data-prevent-default-link-navigation=\"false\" data-link-preview-card-classes=\"MmSINd\"><\/div>\n<\/li>\n<\/ol>\n<\/li>\n<\/ol>\n<\/li>\n<\/ol>\n<\/li>\n<\/ol>\n<p><strong><span class=\"rnc2Gd\" data-ast-node-id=\"39\">3. Feature Extraction &amp; Impulse Pipeline Design<\/span><\/strong><\/p>\n<p><span class=\"rnc2Gd\" data-ast-node-id=\"40\">Once images are collected and labeled, raw pixel values are mapped into numerical feature vectors inside Edge Impuls<\/span><span class=\"rnc2Gd\" data-ast-node-id=\"41\">e.<\/span><\/p>\n<div class=\"gWqWw rnc2Gd\"><\/div>\n<div><img loading=\"lazy\" decoding=\"async\" class=\"alignnone wp-image-39368 size-full\" src=\"\/hazlina\/wp-content\/uploads\/2026\/09\/watermarked_img_8818976253307361420-e1789204063860.jpg\" alt=\"\" width=\"1024\" height=\"547\" srcset=\"\/hazlina\/wp-content\/uploads\/2026\/09\/watermarked_img_8818976253307361420-e1789204063860.jpg 1024w, \/hazlina\/wp-content\/uploads\/2026\/09\/watermarked_img_8818976253307361420-e1789204063860-300x160.jpg 300w, \/hazlina\/wp-content\/uploads\/2026\/09\/watermarked_img_8818976253307361420-e1789204063860-768x410.jpg 768w\" sizes=\"auto, (max-width: 709px) 85vw, (max-width: 909px) 67vw, (max-width: 1362px) 62vw, 840px\" \/><\/div>\n<p><i><span class=\"rnc2Gd\" data-ast-node-id=\"42\">Figure 2: The four-block impulse design connecting input image data to digital signal processing (DSP) and neural network classifier<\/span><span class=\"rnc2Gd\" data-ast-node-id=\"43\">s.<\/span><\/i><\/p>\n<div class=\"rnc2Gd Wv1ane i03hic\" data-open=\"false\" data-type=\"13\" data-url=\"https:\/\/drive.google.com\/file\/d\/1sFTC8DDfPVe-cv68qeAIohGcg3NqKtW6\/view?usp=drive_web\" data-edit-type=\"2\" data-anchor-id=\"ucc-186\" data-anchor-corner=\"5\" data-link-preview-card-corner=\"6\" data-show-thumbnail=\"true\" data-show-preview-delay-ms=\"0\" data-hide-preview-delay-ms=\"0\" data-is-citation-source=\"false\" data-is-internal-link=\"false\" data-show-manually=\"false\" data-animate-entry=\"false\" data-prevent-default-link-navigation=\"false\" data-link-preview-card-classes=\"MmSINd\"><\/div>\n<ol>\n<li style=\"list-style-type: none\">\n<ol>\n<li style=\"list-style-type: none\">\n<ol class=\"gdAMzb\">\n<li class=\"rnc2Gd\"><strong><span class=\"rnc2Gd\" data-ast-node-id=\"44\">Raw Image Preprocessing<\/span><\/strong><span class=\"rnc2Gd\" data-ast-node-id=\"45\">\u00a0Images are normalized and structured into uniform dimension<\/span><span class=\"rnc2Gd\" data-ast-node-id=\"46\">s.<\/span><\/li>\n<li class=\"rnc2Gd\"><strong><span class=\"rnc2Gd\" data-ast-node-id=\"47\">DSP Feature Generation<\/span><\/strong><span class=\"rnc2Gd\" data-ast-node-id=\"48\">\u00a0Spatial characteristics (edges, textures, color distributions) are extracted, generating distinct visual clusters in 3D feature spac<\/span><span class=\"rnc2Gd\" data-ast-node-id=\"49\">e.<\/span><\/li>\n<li class=\"rnc2Gd\"><strong><span class=\"rnc2Gd\" data-ast-node-id=\"50\">Classifier Training<\/span><\/strong><span class=\"rnc2Gd\" data-ast-node-id=\"51\">\u00a0Neural networks process the extracted features to assign probabilities across classe<\/span><span class=\"rnc2Gd\" data-ast-node-id=\"52\">s.<\/span><\/li>\n<li class=\"rnc2Gd\"><strong><span class=\"rnc2Gd\" data-ast-node-id=\"53\">Inference Speed <\/span><\/strong><span class=\"rnc2Gd\" data-ast-node-id=\"54\">Real-time classification executes in approximately\u00a0<\/span><strong><span class=\"rnc2Gd\" data-ast-node-id=\"55\">1 ms<\/span><\/strong><span class=\"rnc2Gd\" data-ast-node-id=\"56\">\u00a0per frame, enabling fluid live monitorin<\/span><span class=\"rnc2Gd\" data-ast-node-id=\"57\">g.<\/span><\/li>\n<\/ol>\n<\/li>\n<\/ol>\n<\/li>\n<\/ol>\n<p><strong><span class=\"rnc2Gd\" data-ast-node-id=\"58\">4. Bounding Boxes, Confidence Thresholds, and the &#8220;Nothing&#8221; Class<\/span><\/strong><\/p>\n<p><span class=\"rnc2Gd\" data-ast-node-id=\"59\">A key challenge in real-world vision deployment is handling unexpected inputs\u2014a scenario highlighted by the &#8220;chili problem<\/span><span class=\"rnc2Gd\" data-ast-node-id=\"60\">&#8220;.<\/span><\/p>\n<div class=\"rnc2Gd Wv1ane i03hic\" data-open=\"false\" data-type=\"13\" data-url=\"https:\/\/drive.google.com\/file\/d\/191HA1cB9vQ-_zSObj55yHeCq781M6VrH\/view?usp=drive_web\" data-edit-type=\"2\" data-anchor-id=\"ucc-191\" data-anchor-corner=\"5\" data-link-preview-card-corner=\"6\" data-show-thumbnail=\"true\" data-show-preview-delay-ms=\"0\" data-hide-preview-delay-ms=\"0\" data-is-citation-source=\"false\" data-is-internal-link=\"false\" data-show-manually=\"false\" data-animate-entry=\"false\" data-prevent-default-link-navigation=\"false\" data-link-preview-card-classes=\"MmSINd\"><img loading=\"lazy\" decoding=\"async\" class=\"alignnone wp-image-39367 size-large\" src=\"\/hazlina\/wp-content\/uploads\/2026\/09\/Screenshot-2026-09-12-at-5.04.54\u202fPM-e1789204179674-1024x578.png\" alt=\"\" width=\"840\" height=\"474\" srcset=\"\/hazlina\/wp-content\/uploads\/2026\/09\/Screenshot-2026-09-12-at-5.04.54\u202fPM-e1789204179674-1024x578.png 1024w, \/hazlina\/wp-content\/uploads\/2026\/09\/Screenshot-2026-09-12-at-5.04.54\u202fPM-e1789204179674-300x169.png 300w, \/hazlina\/wp-content\/uploads\/2026\/09\/Screenshot-2026-09-12-at-5.04.54\u202fPM-e1789204179674-768x434.png 768w, \/hazlina\/wp-content\/uploads\/2026\/09\/Screenshot-2026-09-12-at-5.04.54\u202fPM-e1789204179674-1200x678.png 1200w, \/hazlina\/wp-content\/uploads\/2026\/09\/Screenshot-2026-09-12-at-5.04.54\u202fPM-e1789204179674.png 1725w\" sizes=\"auto, (max-width: 709px) 85vw, (max-width: 909px) 67vw, (max-width: 1362px) 62vw, 840px\" \/><\/div>\n<div data-open=\"false\" data-type=\"13\" data-url=\"https:\/\/drive.google.com\/file\/d\/191HA1cB9vQ-_zSObj55yHeCq781M6VrH\/view?usp=drive_web\" data-edit-type=\"2\" data-anchor-id=\"ucc-191\" data-anchor-corner=\"5\" data-link-preview-card-corner=\"6\" data-show-thumbnail=\"true\" data-show-preview-delay-ms=\"0\" data-hide-preview-delay-ms=\"0\" data-is-citation-source=\"false\" data-is-internal-link=\"false\" data-show-manually=\"false\" data-animate-entry=\"false\" data-prevent-default-link-navigation=\"false\" data-link-preview-card-classes=\"MmSINd\"><\/div>\n<div class=\"rnc2Gd Wv1ane i03hic\" data-open=\"false\" data-type=\"13\" data-url=\"https:\/\/drive.google.com\/file\/d\/1UiYDEBZkU9YliuzKLL1sromhmbPP63f-\/view?usp=drive_web\" data-edit-type=\"2\" data-anchor-id=\"ucc-192\" data-anchor-corner=\"5\" data-link-preview-card-corner=\"6\" data-show-thumbnail=\"true\" data-show-preview-delay-ms=\"0\" data-hide-preview-delay-ms=\"0\" data-is-citation-source=\"false\" data-is-internal-link=\"false\" data-show-manually=\"false\" data-animate-entry=\"false\" data-prevent-default-link-navigation=\"false\" data-link-preview-card-classes=\"MmSINd\"><\/div>\n<div class=\"UrcnEc gM9knf\">\n<table class=\"nU4szd\">\n<thead>\n<tr>\n<th>Challenge<\/th>\n<th>Cause<\/th>\n<th>Embedded Solution<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td><strong><span data-ast-node-id=\"61\">False Positives<\/span><\/strong><\/td>\n<td><span data-ast-node-id=\"62\">The model forces unseen objects into known classe<\/span><span data-ast-node-id=\"63\">s.<\/span><\/td>\n<td><span data-ast-node-id=\"64\">Add an explicit background or &#8220;nothing&#8221; clas<\/span><span data-ast-node-id=\"65\">s.<\/span><\/td>\n<\/tr>\n<tr>\n<td><strong><span data-ast-node-id=\"66\">Fluctuating Predictions<\/span><\/strong><\/td>\n<td><span data-ast-node-id=\"67\">Minor shifts in lighting or tilt alter confidence score<\/span><span data-ast-node-id=\"68\">s.<\/span><\/td>\n<td><span data-ast-node-id=\"69\">Apply confidence thresholds (e.g., ignore results below 0.80<\/span><span data-ast-node-id=\"70\">).<\/span><\/td>\n<\/tr>\n<tr>\n<td><strong><span data-ast-node-id=\"71\">Spatial Localization<\/span><\/strong><\/td>\n<td><span data-ast-node-id=\"72\">Image classification lacks coordinate location.<\/span><\/td>\n<td><span data-ast-node-id=\"73\">Implement bounding box object detection for region trackin<\/span><span data-ast-node-id=\"74\">g.<\/span><\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<div class=\"mjL3ef\" data-generation-index=\"4\" data-session-id=\"goog_224667768\" data-node-id=\"75\" data-is-card=\"false\">\n<div class=\"im7IW\" data-response-buttons-sizing-container=\"true\">\n<div class=\"fnyTGe\" data-generation-index=\"4\" data-session-id=\"goog_224667768\"><\/div>\n<\/div>\n<\/div>\n<p><strong><span class=\"rnc2Gd\" data-ast-node-id=\"76\">5. Microcontroller Integration on ESP32 Hardware<\/span><\/strong><\/p>\n<p><span class=\"rnc2Gd\" data-ast-node-id=\"77\">The culmination of the module is flashing the trained impulse directly onto an ESP microcontroller processo<\/span><span class=\"rnc2Gd\" data-ast-node-id=\"78\">r.<\/span><\/p>\n<div class=\"rnc2Gd Wv1ane i03hic\" data-open=\"false\" data-type=\"13\" data-url=\"https:\/\/drive.google.com\/file\/d\/191HA1cB9vQ-_zSObj55yHeCq781M6VrH\/view?usp=drive_web\" data-edit-type=\"2\" data-anchor-id=\"ucc-202\" data-anchor-corner=\"5\" data-link-preview-card-corner=\"6\" data-show-thumbnail=\"true\" data-show-preview-delay-ms=\"0\" data-hide-preview-delay-ms=\"0\" data-is-citation-source=\"false\" data-is-internal-link=\"false\" data-show-manually=\"false\" data-animate-entry=\"false\" data-prevent-default-link-navigation=\"false\" data-link-preview-card-classes=\"MmSINd\"><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-39370\" src=\"\/hazlina\/wp-content\/uploads\/2026\/09\/watermarked_img_13386078223221394876.jpg\" alt=\"\" width=\"1024\" height=\"1024\" srcset=\"\/hazlina\/wp-content\/uploads\/2026\/09\/watermarked_img_13386078223221394876.jpg 1024w, \/hazlina\/wp-content\/uploads\/2026\/09\/watermarked_img_13386078223221394876-150x150.jpg 150w, \/hazlina\/wp-content\/uploads\/2026\/09\/watermarked_img_13386078223221394876-300x300.jpg 300w, \/hazlina\/wp-content\/uploads\/2026\/09\/watermarked_img_13386078223221394876-768x768.jpg 768w\" sizes=\"auto, (max-width: 709px) 85vw, (max-width: 909px) 67vw, (max-width: 1362px) 62vw, 840px\" \/><\/div>\n<p><i><span class=\"rnc2Gd\" data-ast-node-id=\"79\">Figure 3: Real-time classification outputs streamed over the Arduino Serial Monitor at 115200 bau<\/span><span class=\"rnc2Gd\" data-ast-node-id=\"80\">d.<\/span><\/i><\/p>\n<div class=\"rnc2Gd Wv1ane i03hic\" data-open=\"false\" data-type=\"13\" data-url=\"https:\/\/drive.google.com\/file\/d\/191HA1cB9vQ-_zSObj55yHeCq781M6VrH\/view?usp=drive_web\" data-edit-type=\"2\" data-anchor-id=\"ucc-203\" data-anchor-corner=\"5\" data-link-preview-card-corner=\"6\" data-show-thumbnail=\"true\" data-show-preview-delay-ms=\"0\" data-hide-preview-delay-ms=\"0\" data-is-citation-source=\"false\" data-is-internal-link=\"false\" data-show-manually=\"false\" data-animate-entry=\"false\" data-prevent-default-link-navigation=\"false\" data-link-preview-card-classes=\"MmSINd\"><\/div>\n<p><strong><span class=\"rnc2Gd\" data-ast-node-id=\"81\">Hardware Deployment Steps<\/span><\/strong><\/p>\n<ol>\n<li style=\"list-style-type: none\">\n<ol>\n<li style=\"list-style-type: none\">\n<ol class=\"gdAMzb\">\n<li class=\"rnc2Gd\"><strong><span class=\"rnc2Gd\" data-ast-node-id=\"82\">Firmware Setup<\/span><\/strong><span class=\"rnc2Gd\" data-ast-node-id=\"83\">\u00a0Configure the Arduino IDE with necessary ESP32 board definitions and upload the custom collection sketch (<\/span><code class=\"dfqOJc\">ESP32_datacollection.ino<\/code><span class=\"rnc2Gd\" data-ast-node-id=\"86\">).<\/span><\/li>\n<li class=\"rnc2Gd\"><strong><span class=\"rnc2Gd\" data-ast-node-id=\"87\">Port Selection &amp; Cabling<\/span><\/strong><span class=\"rnc2Gd\" data-ast-node-id=\"88\">\u00a0Ensure high-speed data cables are used, as charge-only cables will fail to establish COM communicatio<\/span><span class=\"rnc2Gd\" data-ast-node-id=\"89\">n.<\/span><\/li>\n<li class=\"rnc2Gd\"><strong><span class=\"rnc2Gd\" data-ast-node-id=\"90\">Network Configurations<\/span><\/strong><span class=\"rnc2Gd\" data-ast-node-id=\"91\">\u00a0Connect via self-generated Wi-Fi access points (Route A) or local network infrastructure (Routes B\/C<\/span><span class=\"rnc2Gd\" data-ast-node-id=\"92\">).<\/span><\/li>\n<li class=\"rnc2Gd\"><strong><span class=\"rnc2Gd\" data-ast-node-id=\"93\">Serial Verification<\/span><\/strong><span class=\"rnc2Gd\" data-ast-node-id=\"94\">\u00a0Initialize the Serial Monitor at\u00a0<\/span><strong><span class=\"rnc2Gd\" data-ast-node-id=\"95\">115200 baud<\/span><\/strong><span class=\"rnc2Gd\" data-ast-node-id=\"96\">\u00a0and press the hardware reset (RST) button to verify initialization and monitor live inference log<\/span><span class=\"rnc2Gd\" data-ast-node-id=\"97\">s.<\/span>\n<div class=\"rnc2Gd Wv1ane i03hic\" data-open=\"false\" data-type=\"13\" data-url=\"https:\/\/drive.google.com\/file\/d\/191HA1cB9vQ-_zSObj55yHeCq781M6VrH\/view?usp=drive_web\" data-edit-type=\"2\" data-anchor-id=\"ucc-207\" data-anchor-corner=\"5\" data-link-preview-card-corner=\"6\" data-show-thumbnail=\"true\" data-show-preview-delay-ms=\"0\" data-hide-preview-delay-ms=\"0\" data-is-citation-source=\"false\" data-is-internal-link=\"false\" data-show-manually=\"false\" data-animate-entry=\"false\" data-prevent-default-link-navigation=\"false\" data-link-preview-card-classes=\"MmSINd\"><\/div>\n<\/li>\n<\/ol>\n<\/li>\n<\/ol>\n<\/li>\n<\/ol>\n<p><strong><span class=\"rnc2Gd\" data-ast-node-id=\"98\">6. Key Takeaways<\/span><\/strong><\/p>\n<p><span class=\"rnc2Gd\" data-ast-node-id=\"99\">Through this module, participants develop end-to-end expertise in embedded machine learning:<\/span><\/p>\n<ol>\n<li style=\"list-style-type: none\">\n<ol class=\"gdAMzb\">\n<li class=\"rnc2Gd\"><span class=\"rnc2Gd\" data-ast-node-id=\"100\">Moving from cloud-dependent AI to standalone micro-processor executio<\/span><span class=\"rnc2Gd\" data-ast-node-id=\"101\">n.<\/span><\/li>\n<li class=\"rnc2Gd\"><span class=\"rnc2Gd\" data-ast-node-id=\"102\">Mastering raw image preprocessing, feature extraction, and dataset hygien<\/span><span class=\"rnc2Gd\" data-ast-node-id=\"103\">e.<\/span><\/li>\n<li class=\"rnc2Gd\"><span class=\"rnc2Gd\" data-ast-node-id=\"104\">Deploying real-time computer vision models on low-cost hardware platform<\/span><span class=\"rnc2Gd\" data-ast-node-id=\"105\">s.<\/span><\/li>\n<\/ol>\n<\/li>\n<\/ol>\n","protected":false},"excerpt":{"rendered":"<p>Deploying artificial intelligence directly on compact microcontrollers\u2014often termed\u00a0Edge Intelligence\u2014represents a major leap in embedded engineering.\u00a0This UMPSA STEM Lab\u00a0module provides a hands-on roadmap for transitioning from basic software-based computer vision to standalone, hardware-integrated AI models capable of performing real-time image processing, feature extraction, and object classification on low-power ESP microcontroller hardware. The Shift to Edge Intelligence &hellip; <a href=\"\/hazlina\/2026\/09\/12\/umpsa-stem-lab-esp32-microcontroller-programming-2026\/\" class=\"more-link\">Continue reading<span class=\"screen-reader-text\"> &#8220;UMPSA STEM Lab &#8211; Edge Intelligence (Image Processing) on ESP32 2026&#8221;<\/span><\/a><\/p>\n","protected":false},"author":2,"featured_media":0,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[2],"tags":[9],"class_list":["post-39360","post","type-post","status-publish","format-standard","hentry","category-information","tag-academia-esd"],"_links":{"self":[{"href":"\/hazlina-blog\/wp-json\/wp\/v2\/posts\/39360","targetHints":{"allow":["GET"]}}],"collection":[{"href":"\/hazlina-blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"\/hazlina-blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"\/hazlina-blog\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"\/hazlina-blog\/wp-json\/wp\/v2\/comments?post=39360"}],"version-history":[{"count":0,"href":"\/hazlina-blog\/wp-json\/wp\/v2\/posts\/39360\/revisions"}],"wp:attachment":[{"href":"\/hazlina-blog\/wp-json\/wp\/v2\/media?parent=39360"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"\/hazlina-blog\/wp-json\/wp\/v2\/categories?post=39360"},{"taxonomy":"post_tag","embeddable":true,"href":"\/hazlina-blog\/wp-json\/wp\/v2\/tags?post=39360"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}