PDF OCR in Kotlin for Android

1
/*
2
 *   Copyright © 2020-2025 PSPDFKit GmbH. All rights reserved.
3
 *
4
 *   The PSPDFKit Sample applications are licensed with a modified BSD license.
5
 *   Please see License for details. This notice may not be removed from this file.
6
 */
7

8
// We're temporarily suppressing the ProgressDialog being deprecated warning.
9
// Issue: https://github.com/PSPDFKit/PSPDFKit/issues/32215
10
@file:Suppress("DEPRECATION")
11

12
package com.pspdfkit.catalog.examples.kotlin
13

14
import android.app.AlertDialog
15
import android.app.ProgressDialog
16
import android.content.Context
17
import android.net.Uri
18
import android.view.Menu
19
import android.view.MenuItem
20
import com.pspdfkit.catalog.R
21
import com.pspdfkit.catalog.SdkExample
22
import com.pspdfkit.catalog.tasks.ExtractAssetTask
23
import com.pspdfkit.configuration.activity.PdfActivityConfiguration
24
import com.pspdfkit.document.processor.PdfProcessor
25
import com.pspdfkit.document.processor.PdfProcessorTask
26
import com.pspdfkit.document.processor.ocr.OcrLanguage
27
import com.pspdfkit.ui.PdfActivity
28
import com.pspdfkit.ui.PdfActivityIntentBuilder
29
import io.reactivex.rxjava3.android.schedulers.AndroidSchedulers
30
import io.reactivex.rxjava3.disposables.CompositeDisposable
31
import io.reactivex.rxjava3.schedulers.Schedulers
32
import io.reactivex.rxjava3.subscribers.DisposableSubscriber
33
import java.io.File
34
import kotlin.math.ceil
35

36
/**
37
 * This examples shows how to use Optical character recognition
38
 * to extract the text from scanned documents using the [com.pspdfkit.document.processor.PdfProcessor].
39
 */
40
class OcrExample(context: Context) : SdkExample(context, R.string.ocrExampleTitle, R.string.ocrExampleDescription) {
41

42
    companion object {
43
        const val OCR_PDF_PATH = "ocr/Remote Work.pdf"
44
    }
45

46
    override fun launchExample(context: Context, configuration: PdfActivityConfiguration.Builder) {
47
        // We use a custom utility class to extract the example document from the assets.
48
        ExtractAssetTask.extract(OCR_PDF_PATH, title, context) { documentFile ->
49
            val intent = PdfActivityIntentBuilder.fromUri(context, Uri.fromFile(documentFile))
50
                .configuration(
51
                    configuration
52
                        .documentInfoViewEnabled(false)
53
                        .annotationListEnabled(false)
54
                        .bookmarkListEnabled(false)
55
                        .settingsMenuEnabled(false)
56
                        .thumbnailGridEnabled(false)
57
                        .outlineEnabled(false).build()
58
                )
59
                .activityClass(OcrProcessingActivity::class)
60
                .build()
61

62
            // Start the OcrProcessingActivity for the extracted document.
63
            context.startActivity(intent)
64
        }
65
    }
66
}
67

68
/**
69
 * This examples shows how to use Optical character recognition
70
 * to extract the text from scanned documents using the [com.pspdfkit.document.processor.PdfProcessor].
71
 */
72
class OcrProcessingActivity : PdfActivity() {
73

74
    private val disposables = CompositeDisposable()
75

76
    /**
77
     * Perform OCR on the current document.
78
     */
79
    private fun performOcr() {
80
        // Define the output file. This example writes to the internal app directory, into a file based on the document's Uid.
81
        val document = document ?: return
82
        val outputFile = File(filesDir, "${document.title}-ocr-processed.pdf")
83
        val pageIndexesToProcess = (0 until document.pageCount).toSet()
84

85
        // Start document processing, perform OCR.
86
        val task = PdfProcessorTask.fromDocument(document)
87
            .performOcrOnPages(pageIndexesToProcess, OcrLanguage.ENGLISH)
88

89
        val handler = ProcessorProgressHandler("Performing OCR on the document.", outputFile)
90

91
        PdfProcessor.processDocumentAsync(task, outputFile)
92
            // Drop update events to avoid back pressure on slow devices.
93
            .onBackpressureDrop()
94
            .subscribeOn(Schedulers.io())
95
            .observeOn(AndroidSchedulers.mainThread())
96
            .doOnCancel { handler.onCancel() }
97
            .subscribe(handler)
98
    }
99

100
    /**
101
     * Creates menu item that will enable document OCR processing.
102
     */
103
    override fun onCreateOptionsMenu(menu: Menu): Boolean {
104
        super.onCreateOptionsMenu(menu)
105
        menuInflater.inflate(R.menu.ocr_example, menu)
106
        return true
107
    }
108

109
    /**
110
     * Triggered by selecting an action from the menu in the action bar.
111
     */
112
    override fun onOptionsItemSelected(item: MenuItem): Boolean {
113
        return when (item.itemId) {
114
            R.id.item_perform_ocr -> {
115
                performOcr()
116
                true
117
            }
118
            else -> super.onOptionsItemSelected(item)
119
        }
120
    }
121

122
    override fun onStop() {
123
        super.onStop()
124
        // Dispose all active disposables when activity goes to background.
125
        disposables.clear()
126
    }
127

128
    private fun showProcessedDocument(processedDocumentFile: File) {
129
        val intent = PdfActivityIntentBuilder.fromUri(this, Uri.fromFile(processedDocumentFile))
130
            .configuration(configuration)
131
            .build()
132
        startActivity(intent)
133
    }
134

135
    /**
136
     * Helper class for showing a progress dialog and opening the processed document.
137
     */
138
    @Suppress("DEPRECATION")
139
    private inner class ProcessorProgressHandler(
140
        progressMessage: String,
141
        private val outputFile: File
142
    ) : DisposableSubscriber<PdfProcessor.ProcessorProgress>() {
143

144
        private val progressDialog: ProgressDialog = ProgressDialog.show(
145
            this@OcrProcessingActivity,
146
            "Processing document",
147
            progressMessage,
148
            false,
149
            true
150
        ) { cancel() }
151

152
        init {
153
            disposables.add(this)
154
        }
155

156
        fun onCancel() {
157
            progressDialog.dismiss()
158
        }
159

160
        override fun onNext(processorProgress: PdfProcessor.ProcessorProgress) {
161
            progressDialog.progress = ceil((processorProgress.pagesProcessed / processorProgress.totalPages).toDouble()).toInt()
162
        }
163

164
        override fun onError(e: Throwable) {
165
            AlertDialog.Builder(this@OcrProcessingActivity)
166
                .setMessage("Error while processing file: " + e.localizedMessage)
167
                .show()
168
            progressDialog.dismiss()
169
        }
170

171
        override fun onComplete() {
172
            showProcessedDocument(outputFile)
173
            progressDialog.dismiss()
174
        }
175
    }
176
}